diff options
Diffstat (limited to 'nixpkgs/doc/stdenv')
| -rw-r--r-- | nixpkgs/doc/stdenv/cross-compilation.chapter.md | 254 | ||||
| -rw-r--r-- | nixpkgs/doc/stdenv/meta.chapter.md | 273 | ||||
| -rw-r--r-- | nixpkgs/doc/stdenv/multiple-output.chapter.md | 128 | ||||
| -rw-r--r-- | nixpkgs/doc/stdenv/platform-notes.chapter.md | 67 | ||||
| -rw-r--r-- | nixpkgs/doc/stdenv/stdenv.chapter.md | 1475 |
5 files changed, 2197 insertions, 0 deletions
diff --git a/nixpkgs/doc/stdenv/cross-compilation.chapter.md b/nixpkgs/doc/stdenv/cross-compilation.chapter.md new file mode 100644 index 000000000000..e659e1803807 --- /dev/null +++ b/nixpkgs/doc/stdenv/cross-compilation.chapter.md @@ -0,0 +1,254 @@ +# Cross-compilation {#chap-cross} + +## Introduction {#sec-cross-intro} + +"Cross-compilation" means compiling a program on one machine for another type of machine. For example, a typical use of cross-compilation is to compile programs for embedded devices. These devices often don't have the computing power and memory to compile their own programs. One might think that cross-compilation is a fairly niche concern. However, there are significant advantages to rigorously distinguishing between build-time and run-time environments! Significant, because the benefits apply even when one is developing and deploying on the same machine. Nixpkgs is increasingly adopting the opinion that packages should be written with cross-compilation in mind, and Nixpkgs should evaluate in a similar way (by minimizing cross-compilation-specific special cases) whether or not one is cross-compiling. + +This chapter will be organized in three parts. First, it will describe the basics of how to package software in a way that supports cross-compilation. Second, it will describe how to use Nixpkgs when cross-compiling. Third, it will describe the internal infrastructure supporting cross-compilation. + +## Packaging in a cross-friendly manner {#sec-cross-packaging} + +### Platform parameters {#ssec-cross-platform-parameters} + +Nixpkgs follows the [conventions of GNU autoconf](https://gcc.gnu.org/onlinedocs/gccint/Configure-Terms.html). We distinguish between 3 types of platforms when building a derivation: _build_, _host_, and _target_. In summary, _build_ is the platform on which a package is being built, _host_ is the platform on which it will run. The third attribute, _target_, is relevant only for certain specific compilers and build tools. + +In Nixpkgs, these three platforms are defined as attribute sets under the names `buildPlatform`, `hostPlatform`, and `targetPlatform`. They are always defined as attributes in the standard environment. That means one can access them like: + +```nix +{ stdenv, fooDep, barDep, ... }: ...stdenv.buildPlatform... +``` + +`buildPlatform` + +: The "build platform" is the platform on which a package is built. Once someone has a built package, or pre-built binary package, the build platform should not matter and can be ignored. + +`hostPlatform` + +: The "host platform" is the platform on which a package will be run. This is the simplest platform to understand, but also the one with the worst name. + +`targetPlatform` + +: The "target platform" attribute is, unlike the other two attributes, not actually fundamental to the process of building software. Instead, it is only relevant for compatibility with building certain specific compilers and build tools. It can be safely ignored for all other packages. + +: The build process of certain compilers is written in such a way that the compiler resulting from a single build can itself only produce binaries for a single platform. The task of specifying this single "target platform" is thus pushed to build time of the compiler. The root cause of this is that the compiler (which will be run on the host) and the standard library/runtime (which will be run on the target) are built by a single build process. + +: There is no fundamental need to think about a single target ahead of time like this. If the tool supports modular or pluggable backends, both the need to specify the target at build time and the constraint of having only a single target disappear. An example of such a tool is LLVM. + +: Although the existence of a "target platform" is arguably a historical mistake, it is a common one: examples of tools that suffer from it are GCC, Binutils, GHC and Autoconf. Nixpkgs tries to avoid sharing in the mistake where possible. Still, because the concept of a target platform is so ingrained, it is best to support it as is. + +The exact schema these fields follow is a bit ill-defined due to a long and convoluted evolution, but this is slowly being cleaned up. You can see examples of ones used in practice in `lib.systems.examples`; note how they are not all very consistent. For now, here are few fields can count on them containing: + +`system` + +: This is a two-component shorthand for the platform. Examples of this would be "x86_64-darwin" and "i686-linux"; see `lib.systems.doubles` for more. The first component corresponds to the CPU architecture of the platform and the second to the operating system of the platform (`[cpu]-[os]`). This format has built-in support in Nix, such as the `builtins.currentSystem` impure string. + +`config` + +: This is a 3- or 4- component shorthand for the platform. Examples of this would be `x86_64-unknown-linux-gnu` and `aarch64-apple-darwin14`. This is a standard format called the "LLVM target triple", as they are pioneered by LLVM. In the 4-part form, this corresponds to `[cpu]-[vendor]-[os]-[abi]`. This format is strictly more informative than the "Nix host double", as the previous format could analogously be termed. This needs a better name than `config`! + +`parsed` + +: This is a Nix representation of a parsed LLVM target triple with white-listed components. This can be specified directly, or actually parsed from the `config`. See `lib.systems.parse` for the exact representation. + +`libc` + +: This is a string identifying the standard C library used. Valid identifiers include "glibc" for GNU libc, "libSystem" for Darwin's Libsystem, and "uclibc" for µClibc. It should probably be refactored to use the module system, like `parse`. + +`is*` + +: These predicates are defined in `lib.systems.inspect`, and slapped onto every platform. They are superior to the ones in `stdenv` as they force the user to be explicit about which platform they are inspecting. Please use these instead of those. + +`platform` + +: This is, quite frankly, a dumping ground of ad-hoc settings (it's an attribute set). See `lib.systems.platforms` for examples—there's hopefully one in there that will work verbatim for each platform that is working. Please help us triage these flags and give them better homes! + +### Theory of dependency categorization {#ssec-cross-dependency-categorization} + +::: {.note} +This is a rather philosophical description that isn't very Nixpkgs-specific. For an overview of all the relevant attributes given to `mkDerivation`, see [](#ssec-stdenv-dependencies). For a description of how everything is implemented, see [](#ssec-cross-dependency-implementation). +::: + +In this section we explore the relationship between both runtime and build-time dependencies and the 3 Autoconf platforms. + +A run time dependency between two packages requires that their host platforms match. This is directly implied by the meaning of "host platform" and "runtime dependency": The package dependency exists while both packages are running on a single host platform. + +A build time dependency, however, has a shift in platforms between the depending package and the depended-on package. "build time dependency" means that to build the depending package we need to be able to run the depended-on's package. The depending package's build platform is therefore equal to the depended-on package's host platform. + +If both the dependency and depending packages aren't compilers or other machine-code-producing tools, we're done. And indeed `buildInputs` and `nativeBuildInputs` have covered these simpler cases for many years. But if the dependency does produce machine code, we might need to worry about its target platform too. In principle, that target platform might be any of the depending package's build, host, or target platforms, but we prohibit dependencies from a "later" platform to an earlier platform to limit confusion because we've never seen a legitimate use for them. + +Finally, if the depending package is a compiler or other machine-code-producing tool, it might need dependencies that run at "emit time". This is for compilers that (regrettably) insist on being built together with their source languages' standard libraries. Assuming build != host != target, a run-time dependency of the standard library cannot be run at the compiler's build time or run time, but only at the run time of code emitted by the compiler. + +Putting this all together, that means that we have dependency types of the form "X→ E", which means that the dependency executes on X and emits code for E; each of X and E can be `build`, `host`, or `target`, and E can be `*` to indicate that the dependency is not a compiler-like package. + +Dependency types describe the relationships that a package has with each of its transitive dependencies. You could think of attaching one or more dependency types to each of the formal parameters at the top of a package's `.nix` file, as well as to all of *their* formal parameters, and so on. Triples like `(foo, bar, baz)`, on the other hand, are a property of an instantiated derivation -- you could would attach a triple `(mips-linux, mips-linux, sparc-solaris)` to a `.drv` file in `/nix/store`. + +Only nine dependency types matter in practice: + +#### Possible dependency types {#possible-dependency-types} + +| Dependency type | Dependency’s host platform | Dependency’s target platform | +|-----------------|----------------------------|------------------------------| +| build → * | build | (none) | +| build → build | build | build | +| build → host | build | host | +| build → target | build | target | +| host → * | host | (none) | +| host → host | host | host | +| host → target | host | target | +| target → * | target | (none) | +| target → target | target | target | + +Let's use `g++` as an example to make this table clearer. `g++` is a C++ compiler written in C. Suppose we are building `g++` with a `(build, host, target)` platform triple of `(foo, bar, baz)`. This means we are using a `foo`-machine to build a copy of `g++` which will run on a `bar`-machine and emit binaries for the `baz`-machine. + +* `g++` links against the host platform's `glibc` C library, which is a "host→ *" dependency with a triple of `(bar, bar, *)`. Since it is a library, not a compiler, it has no "target". + +* Since `g++` is written in C, the `gcc` compiler used to compile it is a "build→ host" dependency of `g++` with a triple of `(foo, foo, bar)`. This compiler runs on the build platform and emits code for the host platform. + +* `gcc` links against the build platform's `glibc` C library, which is a "build→ *" dependency with a triple of `(foo, foo, *)`. Since it is a library, not a compiler, it has no "target". + +* This `gcc` is itself compiled by an *earlier* copy of `gcc`. This earlier copy of `gcc` is a "build→ build" dependency of `g++` with a triple of `(foo, foo, foo)`. This "early `gcc`" runs on the build platform and emits code for the build platform. + +* `g++` is bundled with `libgcc`, which includes a collection of target-machine routines for exception handling and +software floating point emulation. `libgcc` would be a "target→ *" dependency with triple `(foo, baz, *)`, because it consists of machine code which gets linked against the output of the compiler that we are building. It is a library, not a compiler, so it has no target of its own. + +* `libgcc` is written in C and compiled with `gcc`. The `gcc` that compiles it will be a "build→ target" dependency with triple `(foo, foo, baz)`. It gets compiled *and run* at `g++`-build-time (on platform `foo`), but must emit code for the `baz`-platform. + +* `g++` allows inline assembler code, so it depends on access to a copy of the `gas` assembler. This would be a "host→ target" dependency with triple `(foo, bar, baz)`. + +* `g++` (and `gcc`) include a library `libgccjit.so`, which wrap the compiler in a library to create a just-in-time compiler. In nixpkgs, this library is in the `libgccjit` package; if C++ required that programs have access to a JIT, `g++` would need to add a "target→ target" dependency for `libgccjit` with triple `(foo, baz, baz)`. This would ensure that the compiler ships with a copy of `libgccjit` which both executes on and generates code for the `baz`-platform. + +* If `g++` itself linked against `libgccjit.so` (for example, to allow compile-time-evaluated C++ expressions), then the `libgccjit` package used to provide this functionality would be a "host→ host" dependency of `g++`: it is code which runs on the `host` and emits code for execution on the `host`. + +### Cross packaging cookbook {#ssec-cross-cookbook} + +Some frequently encountered problems when packaging for cross-compilation should be answered here. Ideally, the information above is exhaustive, so this section cannot provide any new information, but it is ludicrous and cruel to expect everyone to spend effort working through the interaction of many features just to figure out the same answer to the same common problem. Feel free to add to this list! + +#### My package fails to find a binutils command (`cc`/`ar`/`ld` etc.) {#cross-qa-fails-to-find-binutils} +Many packages assume that an unprefixed binutils (`cc`/`ar`/`ld` etc.) is available, but Nix doesn't provide one. It only provides a prefixed one, just as it only does for all the other binutils programs. It may be necessary to patch the package to fix the build system to use a prefix. For instance, instead of `cc`, use `${stdenv.cc.targetPrefix}cc`. + +```nix +makeFlags = [ "CC=${stdenv.cc.targetPrefix}cc" ]; +``` + +#### How do I avoid compiling a GCC cross-compiler from source? {#cross-qa-avoid-compiling-gcc-cross-compiler} +On less powerful machines, it can be inconvenient to cross-compile a package only to find out that GCC has to be compiled from source, which could take up to several hours. Nixpkgs maintains a limited [cross-related jobset on Hydra](https://hydra.nixos.org/jobset/nixpkgs/cross-trunk), which tests cross-compilation to various platforms from build platforms "x86\_64-darwin", "x86\_64-linux", and "aarch64-linux". See `pkgs/top-level/release-cross.nix` for the full list of target platforms and packages. For instance, the following invocation fetches the pre-built cross-compiled GCC for `armv6l-unknown-linux-gnueabihf` and builds GNU Hello from source. + +```ShellSession +$ nix-build '<nixpkgs>' -A pkgsCross.raspberryPi.hello +``` + +#### What if my package’s build system needs to build a C program to be run under the build environment? {#cross-qa-build-c-program-in-build-environment} + +Add the following to your `mkDerivation` invocation. + +```nix +depsBuildBuild = [ buildPackages.stdenv.cc ]; +``` + +#### My package’s testsuite needs to run host platform code. {#cross-testsuite-runs-host-code} + +Add the following to your `mkDerivation` invocation. + +```nix +doCheck = stdenv.buildPlatform.canExecute stdenv.hostPlatform; +``` + +#### Package using Meson needs to run binaries for the host platform during build. {#cross-meson-runs-host-code} + +Add `mesonEmulatorHook` to `nativeBuildInputs` conditionally on if the target binaries can be executed. + +e.g. + +``` +nativeBuildInputs = [ + meson +] ++ lib.optionals (!stdenv.buildPlatform.canExecute stdenv.hostPlatform) [ + mesonEmulatorHook +]; +``` + +Example of an error which this fixes. + +`[Errno 8] Exec format error: './gdk3-scan'` + +## Cross-building packages {#sec-cross-usage} + +Nixpkgs can be instantiated with `localSystem` alone, in which case there is no cross-compiling and everything is built by and for that system, or also with `crossSystem`, in which case packages run on the latter, but all building happens on the former. Both parameters take the same schema as the 3 (build, host, and target) platforms defined in the previous section. As mentioned above, `lib.systems.examples` has some platforms which are used as arguments for these parameters in practice. You can use them programmatically, or on the command line: + +```ShellSession +$ nix-build '<nixpkgs>' --arg crossSystem '(import <nixpkgs/lib>).systems.examples.fooBarBaz' -A whatever +``` + +::: {.note} +Eventually we would like to make these platform examples an unnecessary convenience so that + +```ShellSession +$ nix-build '<nixpkgs>' --arg crossSystem '{ config = "<arch>-<os>-<vendor>-<abi>"; }' -A whatever +``` + +works in the vast majority of cases. The problem today is dependencies on other sorts of configuration which aren't given proper defaults. We rely on the examples to crudely to set those configuration parameters in some vaguely sane manner on the users behalf. Issue [\#34274](https://github.com/NixOS/nixpkgs/issues/34274) tracks this inconvenience along with its root cause in crufty configuration options. +::: + +While one is free to pass both parameters in full, there's a lot of logic to fill in missing fields. As discussed in the previous section, only one of `system`, `config`, and `parsed` is needed to infer the other two. Additionally, `libc` will be inferred from `parse`. Finally, `localSystem.system` is also _impurely_ inferred based on the platform evaluation occurs. This means it is often not necessary to pass `localSystem` at all, as in the command-line example in the previous paragraph. + +::: {.note} +Many sources (manual, wiki, etc) probably mention passing `system`, `platform`, along with the optional `crossSystem` to Nixpkgs: `import <nixpkgs> { system = ..; platform = ..; crossSystem = ..; }`. Passing those two instead of `localSystem` is still supported for compatibility, but is discouraged. Indeed, much of the inference we do for these parameters is motivated by compatibility as much as convenience. +::: + +One would think that `localSystem` and `crossSystem` overlap horribly with the three `*Platforms` (`buildPlatform`, `hostPlatform,` and `targetPlatform`; see `stage.nix` or the manual). Actually, those identifiers are purposefully not used here to draw a subtle but important distinction: While the granularity of having 3 platforms is necessary to properly *build* packages, it is overkill for specifying the user's *intent* when making a build plan or package set. A simple "build vs deploy" dichotomy is adequate: the sliding window principle described in the previous section shows how to interpolate between the these two "end points" to get the 3 platform triple for each bootstrapping stage. That means for any package a given package set, even those not bound on the top level but only reachable via dependencies or `buildPackages`, the three platforms will be defined as one of `localSystem` or `crossSystem`, with the former replacing the latter as one traverses build-time dependencies. A last simple difference is that `crossSystem` should be null when one doesn't want to cross-compile, while the `*Platform`s are always non-null. `localSystem` is always non-null. + +## Cross-compilation infrastructure {#sec-cross-infra} + +### Implementation of dependencies {#ssec-cross-dependency-implementation} + +The categories of dependencies developed in [](#ssec-cross-dependency-categorization) are specified as lists of derivations given to `mkDerivation`, as documented in [](#ssec-stdenv-dependencies). In short, each list of dependencies for "host → target" is called `deps<host><target>` (where `host`, and `target` values are either `build`, `host`, or `target`), with exceptions for backwards compatibility that `depsBuildHost` is instead called `nativeBuildInputs` and `depsHostTarget` is instead called `buildInputs`. Nixpkgs is now structured so that each `deps<host><target>` is automatically taken from `pkgs<host><target>`. (These `pkgs<host><target>`s are quite new, so there is no special case for `nativeBuildInputs` and `buildInputs`.) For example, `pkgsBuildHost.gcc` should be used at build-time, while `pkgsHostTarget.gcc` should be used at run-time. + +Now, for most of Nixpkgs's history, there were no `pkgs<host><target>` attributes, and most packages have not been refactored to use it explicitly. Prior to those, there were just `buildPackages`, `pkgs`, and `targetPackages`. Those are now redefined as aliases to `pkgsBuildHost`, `pkgsHostTarget`, and `pkgsTargetTarget`. It is acceptable, even recommended, to use them for libraries to show that the host platform is irrelevant. + +But before that, there was just `pkgs`, even though both `buildInputs` and `nativeBuildInputs` existed. \[Cross barely worked, and those were implemented with some hacks on `mkDerivation` to override dependencies.\] What this means is the vast majority of packages do not use any explicit package set to populate their dependencies, just using whatever `callPackage` gives them even if they do correctly sort their dependencies into the multiple lists described above. And indeed, asking that users both sort their dependencies, _and_ take them from the right attribute set, is both too onerous and redundant, so the recommended approach (for now) is to continue just categorizing by list and not using an explicit package set. + +To make this work, we "splice" together the six `pkgsFooBar` package sets and have `callPackage` actually take its arguments from that. This is currently implemented in `pkgs/top-level/splice.nix`. `mkDerivation` then, for each dependency attribute, pulls the right derivation out from the splice. This splicing can be skipped when not cross-compiling as the package sets are the same, but still is a bit slow for cross-compiling. We'd like to do something better, but haven't come up with anything yet. + +### Bootstrapping {#ssec-bootstrapping} + +Each of the package sets described above come from a single bootstrapping stage. While `pkgs/top-level/default.nix`, coordinates the composition of stages at a high level, `pkgs/top-level/stage.nix` "ties the knot" (creates the fixed point) of each stage. The package sets are defined per-stage however, so they can be thought of as edges between stages (the nodes) in a graph. Compositions like `pkgsBuildTarget.targetPackages` can be thought of as paths to this graph. + +While there are many package sets, and thus many edges, the stages can also be arranged in a linear chain. In other words, many of the edges are redundant as far as connectivity is concerned. This hinges on the type of bootstrapping we do. Currently for cross it is: + +1. `(native, native, native)` + +2. `(native, native, foreign)` + +3. `(native, foreign, foreign)` + +In each stage, `pkgsBuildHost` refers to the previous stage, `pkgsBuildBuild` refers to the one before that, and `pkgsHostTarget` refers to the current one, and `pkgsTargetTarget` refers to the next one. When there is no previous or next stage, they instead refer to the current stage. Note how all the invariants regarding the mapping between dependency and depending packages' build host and target platforms are preserved. `pkgsBuildTarget` and `pkgsHostHost` are more complex in that the stage fitting the requirements isn't always a fixed chain of "prevs" and "nexts" away (modulo the "saturating" self-references at the ends). We just special case each instead. All the primary edges are implemented is in `pkgs/stdenv/booter.nix`, and secondarily aliases in `pkgs/top-level/stage.nix`. + +::: {.note} +The native stages are bootstrapped in legacy ways that predate the current cross implementation. This is why the bootstrapping stages leading up to the final stages are ignored in the previous paragraph. +::: + +If one looks at the 3 platform triples, one can see that they overlap such that one could put them together into a chain like: +``` +(native, native, native, foreign, foreign) +``` + +If one imagines the saturating self references at the end being replaced with infinite stages, and then overlays those platform triples, one ends up with the infinite tuple: +``` +(native..., native, native, native, foreign, foreign, foreign...) +``` +One can then imagine any sequence of platforms such that there are bootstrap stages with their 3 platforms determined by "sliding a window" that is the 3 tuple through the sequence. This was the original model for bootstrapping. Without a target platform (assume a better world where all compilers are multi-target and all standard libraries are built in their own derivation), this is sufficient. Conversely if one wishes to cross compile "faster", with a "Canadian Cross" bootstrapping stage where `build != host != target`, more bootstrapping stages are needed since no sliding window provides the pesky `pkgsBuildTarget` package set since it skips the Canadian cross stage's "host". + + +::: {.note} +It is much better to refer to `buildPackages` than `targetPackages`, or more broadly package sets that do not mention “target”. There are three reasons for this. + +First, it is because bootstrapping stages do not have a unique `targetPackages`. For example a `(x86-linux, x86-linux, arm-linux)` and `(x86-linux, x86-linux, x86-windows)` package set both have a `(x86-linux, x86-linux, x86-linux)` package set. Because there is no canonical `targetPackages` for such a native (`build == host == target`) package set, we set their `targetPackages` + +Second, it is because this is a frequent source of hard-to-follow "infinite recursions" / cycles. When only package sets that don't mention target are used, the package set forms a directed acyclic graph. This means that all cycles that exist are confined to one stage. This means they are a lot smaller, and easier to follow in the code or a backtrace. It also means they are present in native and cross builds alike, and so more likely to be caught by CI and other users. + +Thirdly, it is because everything target-mentioning only exists to accommodate compilers with lousy build systems that insist on the compiler itself and standard library being built together. Of course that is bad because bigger derivations means longer rebuilds. It is also problematic because it tends to make the standard libraries less like other libraries than they could be, complicating code and build systems alike. Because of the other problems, and because of these innate disadvantages, compilers ought to be packaged another way where possible. +::: + +::: {.note} +If one explores Nixpkgs, they will see derivations with names like `gccCross`. Such `*Cross` derivations is a holdover from before we properly distinguished between the host and target platforms—the derivation with “Cross” in the name covered the `build = host != target` case, while the other covered the `host = target`, with build platform the same or not based on whether one was using its `.__spliced.buildHost` or `.__spliced.hostTarget`. +::: diff --git a/nixpkgs/doc/stdenv/meta.chapter.md b/nixpkgs/doc/stdenv/meta.chapter.md new file mode 100644 index 000000000000..0cb2d6573dfc --- /dev/null +++ b/nixpkgs/doc/stdenv/meta.chapter.md @@ -0,0 +1,273 @@ +# Meta-attributes {#chap-meta} + +Nix packages can declare *meta-attributes* that contain information about a package such as a description, its homepage, its license, and so on. For instance, the GNU Hello package has a `meta` declaration like this: + +```nix +meta = with lib; { + description = "A program that produces a familiar, friendly greeting"; + longDescription = '' + GNU Hello is a program that prints "Hello, world!" when you run it. + It is fully customizable. + ''; + homepage = "https://www.gnu.org/software/hello/manual/"; + license = licenses.gpl3Plus; + maintainers = with maintainers; [ eelco ]; + platforms = platforms.all; +}; +``` + +Meta-attributes are not passed to the builder of the package. Thus, a change to a meta-attribute doesn’t trigger a recompilation of the package. + +## Standard meta-attributes {#sec-standard-meta-attributes} + +It is expected that each meta-attribute is one of the following: + +### `description` {#var-meta-description} + +A short (one-line) description of the package. This is shown by `nix-env -q --description` and also on the Nixpkgs release pages. + +Don’t include a period at the end. Don’t include newline characters. Capitalise the first character. For brevity, don’t repeat the name of package --- just describe what it does. + +Wrong: `"libpng is a library that allows you to decode PNG images."` + +Right: `"A library for decoding PNG images"` + +### `longDescription` {#var-meta-longDescription} + +An arbitrarily long description of the package in [CommonMark](https://commonmark.org) Markdown. + +### `branch` {#var-meta-branch} + +Release branch. Used to specify that a package is not going to receive updates that are not in this branch; for example, Linux kernel 3.0 is supposed to be updated to 3.0.X, not 3.1. + +### `homepage` {#var-meta-homepage} + +The package’s homepage. Example: `https://www.gnu.org/software/hello/manual/` + +### `downloadPage` {#var-meta-downloadPage} + +The page where a link to the current version can be found. Example: `https://ftp.gnu.org/gnu/hello/` + +### `changelog` {#var-meta-changelog} + +A link or a list of links to the location of Changelog for a package. A link may use expansion to refer to the correct changelog version. Example: `"https://git.savannah.gnu.org/cgit/hello.git/plain/NEWS?h=v${version}"` + +### `license` {#var-meta-license} + +The license, or licenses, for the package. One from the attribute set defined in [`nixpkgs/lib/licenses.nix`](https://github.com/NixOS/nixpkgs/blob/master/lib/licenses.nix). At this moment using both a list of licenses and a single license is valid. If the license field is in the form of a list representation, then it means that parts of the package are licensed differently. Each license should preferably be referenced by their attribute. The non-list attribute value can also be a space delimited string representation of the contained attribute `shortNames` or `spdxIds`. The following are all valid examples: + +- Single license referenced by attribute (preferred) `lib.licenses.gpl3Only`. +- Single license referenced by its attribute shortName (frowned upon) `"gpl3Only"`. +- Single license referenced by its attribute spdxId (frowned upon) `"GPL-3.0-only"`. +- Multiple licenses referenced by attribute (preferred) `with lib.licenses; [ asl20 free ofl ]`. +- Multiple licenses referenced as a space delimited string of attribute shortNames (frowned upon) `"asl20 free ofl"`. + +For details, see [Licenses](#sec-meta-license). + +### `maintainers` {#var-meta-maintainers} + +A list of the maintainers of this Nix expression. Maintainers are defined in [`nixpkgs/maintainers/maintainer-list.nix`](https://github.com/NixOS/nixpkgs/blob/master/maintainers/maintainer-list.nix). There is no restriction to becoming a maintainer, just add yourself to that list in a separate commit titled “maintainers: add alice” in the same pull request, and reference maintainers with `maintainers = with lib.maintainers; [ alice bob ]`. + +### `mainProgram` {#var-meta-mainProgram} + +The name of the main binary for the package. This affects the binary `nix run` executes and falls back to the name of the package. Example: `"rg"` + +### `priority` {#var-meta-priority} + +The *priority* of the package, used by `nix-env` to resolve file name conflicts between packages. See the Nix manual page for `nix-env` for details. Example: `"10"` (a low-priority package). + +### `platforms` {#var-meta-platforms} + +The list of Nix platform types on which the package is supported. Hydra builds packages according to the platform specified. If no platform is specified, the package does not have prebuilt binaries. An example is: + +```nix +meta.platforms = lib.platforms.linux; +``` + +Attribute Set `lib.platforms` defines [various common lists](https://github.com/NixOS/nixpkgs/blob/master/lib/systems/doubles.nix) of platforms types. + +### `badPlatforms` {#var-meta-badPlatforms} + +The list of Nix [platform types](https://github.com/NixOS/nixpkgs/blob/b03ac42b0734da3e7be9bf8d94433a5195734b19/lib/meta.nix#L75-L81) on which the package is known not to be buildable. +Hydra will never create prebuilt binaries for these platform types, even if they are in [`meta.platforms`](#var-meta-platforms). +In general it is preferable to set `meta.platforms = lib.platforms.all` and then exclude any platforms on which the package is known not to build. +For example, a package which requires dynamic linking and cannot be linked statically could use this: + +```nix +meta.platforms = lib.platforms.all; +meta.badPlatforms = [ lib.systems.inspect.patterns.isStatic ]; +``` + +The [`lib.meta.availableOn`](https://github.com/NixOS/nixpkgs/blob/b03ac42b0734da3e7be9bf8d94433a5195734b19/lib/meta.nix#L95-L106) function can be used to test whether or not a package is available (i.e. buildable) on a given platform. +Some packages use this to automatically detect the maximum set of features with which they can be built. +For example, `systemd` [requires dynamic linking](https://github.com/systemd/systemd/issues/20600#issuecomment-912338965), and [has a `meta.badPlatforms` setting](https://github.com/NixOS/nixpkgs/blob/b03ac42b0734da3e7be9bf8d94433a5195734b19/pkgs/os-specific/linux/systemd/default.nix#L752) similar to the one above. +Packages which can be built with or without `systemd` support will use `lib.meta.availableOn` to detect whether or not `systemd` is available on the [`hostPlatform`](#ssec-cross-platform-parameters) for which they are being built; if it is not available (e.g. due to a statically-linked host platform like `pkgsStatic`) this support will be disabled by default. + +### `tests` {#var-meta-tests} + +::: {.warning} +This attribute is special in that it is not actually under the `meta` attribute set but rather under the `passthru` attribute set. This is due to how `meta` attributes work, and the fact that they are supposed to contain only metadata, not derivations. +::: + +An attribute set with tests as values. A test is a derivation that builds when the test passes and fails to build otherwise. + +You can run these tests with: + +```ShellSession +$ cd path/to/nixpkgs +$ nix-build -A your-package.tests +``` + +#### Package tests {#var-meta-tests-packages} + +Tests that are part of the source package are often executed in the `installCheckPhase`. + +Prefer `passthru.tests` for tests that are introduced in nixpkgs because: + +* `passthru.tests` tests the 'real' package, independently from the environment in which it was built +* we can run `passthru.tests` independently +* `installCheckPhase` adds overhead to each build + +For more on how to write and run package tests, see <xref linkend="sec-package-tests"/>. + +#### NixOS tests {#var-meta-tests-nixos} + +The NixOS tests are available as `nixosTests` in parameters of derivations. For instance, the OpenSMTPD derivation includes lines similar to: + +```nix +{ /* ... */, nixosTests }: +{ + # ... + passthru.tests = { + basic-functionality-and-dovecot-integration = nixosTests.opensmtpd; + }; +} +``` + +NixOS tests run in a VM, so they are slower than regular package tests. For more information see [NixOS module tests](https://nixos.org/manual/nixos/stable/#sec-nixos-tests). + +Alternatively, you can specify other derivations as tests. You can make use of +the optional parameter to inject the correct package without +relying on non-local definitions, even in the presence of `overrideAttrs`. +Here that's `finalAttrs.finalPackage`, but you could choose a different name if +`finalAttrs` already exists in your scope. + +`(mypkg.overrideAttrs f).passthru.tests` will be as expected, as long as the +definition of `tests` does not rely on the original `mypkg` or overrides it in +all places. + +```nix +# my-package/default.nix +{ stdenv, callPackage }: +stdenv.mkDerivation (finalAttrs: { + # ... + passthru.tests.example = callPackage ./example.nix { my-package = finalAttrs.finalPackage; }; +}) +``` + +```nix +# my-package/example.nix +{ runCommand, lib, my-package, ... }: +runCommand "my-package-test" { + nativeBuildInputs = [ my-package ]; + src = lib.sources.sourcesByRegex ./. [ ".*.in" ".*.expected" ]; +} '' + my-package --help + my-package <example.in >example.actual + diff -U3 --color=auto example.expected example.actual + mkdir $out +'' +``` + + +### `timeout` {#var-meta-timeout} + +A timeout (in seconds) for building the derivation. If the derivation takes longer than this time to build, Hydra will fail it due to breaking the timeout. However, all computers do not have the same computing power, hence some builders may decide to apply a multiplicative factor to this value. When filling this value in, try to keep it approximately consistent with other values already present in `nixpkgs`. + +`meta` attributes are not stored in the instantiated derivation. +Therefore, this setting may be lost when the package is used as a dependency. +To be effective, it must be presented directly to an evaluation process that handles the `meta.timeout` attribute. + +### `hydraPlatforms` {#var-meta-hydraPlatforms} + +The list of Nix platform types for which the [Hydra](https://github.com/nixos/hydra) [instance at `hydra.nixos.org`](https://nixos.org/hydra) will build the package. (Hydra is the Nix-based continuous build system.) It defaults to the value of `meta.platforms`. Thus, the only reason to set `meta.hydraPlatforms` is if you want `hydra.nixos.org` to build the package on a subset of `meta.platforms`, or not at all, e.g. + +```nix +meta.platforms = lib.platforms.linux; +meta.hydraPlatforms = []; +``` + +### `broken` {#var-meta-broken} + +If set to `true`, the package is marked as "broken", meaning that it won’t show up in [search.nixos.org](https://search.nixos.org/packages), and cannot be built or installed unless the environment variable [`NIXPKGS_ALLOW_BROKEN`](#opt-allowBroken) is set. +Such unconditionally-broken packages should be removed from Nixpkgs eventually unless they are fixed. + +The value of this attribute can depend on a package's arguments, including `stdenv`. +This means that `broken` can be used to express constraints, for example: + +- Does not cross compile + + ```nix + meta.broken = !(stdenv.buildPlatform.canExecute stdenv.hostPlatform) + ``` + +- Broken if all of a certain set of its dependencies are broken + + ```nix + meta.broken = lib.all (map (p: p.meta.broken) [ glibc musl ]) + ``` + +This makes `broken` strictly more powerful than `meta.badPlatforms`. +However `meta.availableOn` currently examines only `meta.platforms` and `meta.badPlatforms`, so `meta.broken` does not influence the default values for optional dependencies. + +## Licenses {#sec-meta-license} + +The `meta.license` attribute should preferably contain a value from `lib.licenses` defined in [`nixpkgs/lib/licenses.nix`](https://github.com/NixOS/nixpkgs/blob/master/lib/licenses.nix), or in-place license description of the same format if the license is unlikely to be useful in another expression. + +Although it’s typically better to indicate the specific license, a few generic options are available: + +### `lib.licenses.free`, `"free"` {#lib.licenses.free-free} + +Catch-all for free software licenses not listed above. + +### `lib.licenses.unfreeRedistributable`, `"unfree-redistributable"` {#lib.licenses.unfreeredistributable-unfree-redistributable} + +Unfree package that can be redistributed in binary form. That is, it’s legal to redistribute the *output* of the derivation. This means that the package can be included in the Nixpkgs channel. + +Sometimes proprietary software can only be redistributed unmodified. Make sure the builder doesn’t actually modify the original binaries; otherwise we’re breaking the license. For instance, the NVIDIA X11 drivers can be redistributed unmodified, but our builder applies `patchelf` to make them work. Thus, its license is `"unfree"` and it cannot be included in the Nixpkgs channel. + +### `lib.licenses.unfree`, `"unfree"` {#lib.licenses.unfree-unfree} + +Unfree package that cannot be redistributed. You can build it yourself, but you cannot redistribute the output of the derivation. Thus it cannot be included in the Nixpkgs channel. + +### `lib.licenses.unfreeRedistributableFirmware`, `"unfree-redistributable-firmware"` {#lib.licenses.unfreeredistributablefirmware-unfree-redistributable-firmware} + +This package supplies unfree, redistributable firmware. This is a separate value from `unfree-redistributable` because not everybody cares whether firmware is free. + +## Source provenance {#sec-meta-sourceProvenance} + +The value of a package's `meta.sourceProvenance` attribute specifies the provenance of the package's derivation outputs. + +If a package contains elements that are not built from the original source by a nixpkgs derivation, the `meta.sourceProvenance` attribute should be a list containing one or more value from `lib.sourceTypes` defined in [`nixpkgs/lib/source-types.nix`](https://github.com/NixOS/nixpkgs/blob/master/lib/source-types.nix). + +Adding this information helps users who have needs related to build transparency and supply-chain security to gain some visibility into their installed software or set policy to allow or disallow installation based on source provenance. + +The presence of a particular `sourceType` in a package's `meta.sourceProvenance` list indicates that the package contains some components falling into that category, though the *absence* of that `sourceType` does not *guarantee* the absence of that category of `sourceType` in the package's contents. A package with no `meta.sourceProvenance` set implies it has no *known* `sourceType`s other than `fromSource`. + +The meaning of the `meta.sourceProvenance` attribute does not depend on the value of the `meta.license` attribute. + +### `lib.sourceTypes.fromSource` {#lib.sourceTypes.fromSource} + +Package elements which are produced by a nixpkgs derivation which builds them from source code. + +### `lib.sourceTypes.binaryNativeCode` {#lib.sourceTypes.binaryNativeCode} + +Native code to be executed on the target system's CPU, built by a third party. This includes packages which wrap a downloaded AppImage or Debian package. + +### `lib.sourceTypes.binaryFirmware` {#lib.sourceTypes.binaryFirmware} + +Code to be executed on a peripheral device or embedded controller, built by a third party. + +### `lib.sourceTypes.binaryBytecode` {#lib.sourceTypes.binaryBytecode} + +Code to run on a VM interpreter or JIT compiled into bytecode by a third party. This includes packages which download Java `.jar` files from another source. diff --git a/nixpkgs/doc/stdenv/multiple-output.chapter.md b/nixpkgs/doc/stdenv/multiple-output.chapter.md new file mode 100644 index 000000000000..c19d497ab61e --- /dev/null +++ b/nixpkgs/doc/stdenv/multiple-output.chapter.md @@ -0,0 +1,128 @@ +# Multiple-output packages {#chap-multiple-output} + +## Introduction {#sec-multiple-outputs-introduction} + +The Nix language allows a derivation to produce multiple outputs, which is similar to what is utilized by other Linux distribution packaging systems. The outputs reside in separate Nix store paths, so they can be mostly handled independently of each other, including passing to build inputs, garbage collection or binary substitution. The exception is that building from source always produces all the outputs. + +The main motivation is to save disk space by reducing runtime closure sizes; consequently also sizes of substituted binaries get reduced. Splitting can be used to have more granular runtime dependencies, for example the typical reduction is to split away development-only files, as those are typically not needed during runtime. As a result, closure sizes of many packages can get reduced to a half or even much less. + +::: {.note} +The reduction effects could be instead achieved by building the parts in completely separate derivations. That would often additionally reduce build-time closures, but it tends to be much harder to write such derivations, as build systems typically assume all parts are being built at once. This compromise approach of single source package producing multiple binary packages is also utilized often by rpm and deb. +::: + +A number of attributes can be used to work with a derivation with multiple outputs. The attribute `outputs` is a list of strings, which are the names of the outputs. For each of these names, an identically named attribute is created, corresponding to that output. The attribute `meta.outputsToInstall` is used to determine the default set of outputs to install when using the derivation name unqualified. + +## Installing a split package {#sec-multiple-outputs-installing} + +When installing a package with multiple outputs, the package’s `meta.outputsToInstall` attribute determines which outputs are actually installed. `meta.outputsToInstall` is a list whose [default installs binaries and the associated man pages](https://github.com/NixOS/nixpkgs/blob/f1680774340d5443a1409c3421ced84ac1163ba9/pkgs/stdenv/generic/make-derivation.nix#L310-L320). The following sections describe ways to install different outputs. + +### Selecting outputs to install via NixOS {#sec-multiple-outputs-installing-nixos} + +NixOS provides two ways to select the outputs to install for packages listed in `environment.systemPackages`: + +- The configuration option `environment.extraOutputsToInstall` is appended to each package’s `meta.outputsToInstall` attribute to determine the outputs to install. It can for example be used to install `info` documentation or debug symbols for all packages. + +- The outputs can be listed as packages in `environment.systemPackages`. For example, the `"out"` and `"info"` outputs for the `coreutils` package can be installed by including `coreutils` and `coreutils.info` in `environment.systemPackages`. + +### Selecting outputs to install via `nix-env` {#sec-multiple-outputs-installing-nix-env} + +`nix-env` lacks an easy way to select the outputs to install. When installing a package, `nix-env` always installs the outputs listed in `meta.outputsToInstall`, even when the user explicitly selects an output. + +::: {.warning} +`nix-env` silently disregards the outputs selected by the user, and instead installs the outputs from `meta.outputsToInstall`. For example, + +```ShellSession +$ nix-env -iA nixpkgs.coreutils.info +``` + +installs the `"out"` output (`coreutils.meta.outputsToInstall` is `[ "out" ]`) instead of the requested `"info"`. +::: + +The only recourse to select an output with `nix-env` is to override the package’s `meta.outputsToInstall`, using the functions described in [](#chap-overrides). For example, the following overlay adds the `"info"` output for the `coreutils` package: + +```nix +self: super: +{ + coreutils = super.coreutils.overrideAttrs (oldAttrs: { + meta = oldAttrs.meta // { outputsToInstall = oldAttrs.meta.outputsToInstall or [ "out" ] ++ [ "info" ]; }; + }); +} +``` + +## Using a split package {#sec-multiple-outputs-using-split-packages} + +In the Nix language the individual outputs can be reached explicitly as attributes, e.g. `coreutils.info`, but the typical case is just using packages as build inputs. + +When a multiple-output derivation gets into a build input of another derivation, the `dev` output is added if it exists, otherwise the first output is added. In addition to that, `propagatedBuildOutputs` of that package which by default contain `$outputBin` and `$outputLib` are also added. (See [](#multiple-output-file-type-groups).) + +In some cases it may be desirable to combine different outputs under a single store path. A function `symlinkJoin` can be used to do this. (Note that it may negate some closure size benefits of using a multiple-output package.) + +## Writing a split derivation {#sec-multiple-outputs-} + +Here you find how to write a derivation that produces multiple outputs. + +In nixpkgs there is a framework supporting multiple-output derivations. It tries to cover most cases by default behavior. You can find the source separated in `<nixpkgs/pkgs/build-support/setup-hooks/multiple-outputs.sh>`; it’s relatively well-readable. The whole machinery is triggered by defining the `outputs` attribute to contain the list of desired output names (strings). + +```nix +outputs = [ "bin" "dev" "out" "doc" ]; +``` + +Often such a single line is enough. For each output an equally named environment variable is passed to the builder and contains the path in nix store for that output. Typically you also want to have the main `out` output, as it catches any files that didn’t get elsewhere. + +::: {.note} +There is a special handling of the `debug` output, described at [](#stdenv-separateDebugInfo). +::: + +### “Binaries first” {#multiple-output-file-binaries-first-convention} + +A commonly adopted convention in `nixpkgs` is that executables provided by the package are contained within its first output. This convention allows the dependent packages to reference the executables provided by packages in a uniform manner. For instance, provided with the knowledge that the `perl` package contains a `perl` executable it can be referenced as `${pkgs.perl}/bin/perl` within a Nix derivation that needs to execute a Perl script. + +The `glibc` package is a deliberate single exception to the “binaries first” convention. The `glibc` has `libs` as its first output allowing the libraries provided by `glibc` to be referenced directly (e.g. `${glibc}/lib/ld-linux-x86-64.so.2`). The executables provided by `glibc` can be accessed via its `bin` attribute (e.g. `${lib.getBin stdenv.cc.libc}/bin/ldd`). + +The reason for why `glibc` deviates from the convention is because referencing a library provided by `glibc` is a very common operation among Nix packages. For instance, third-party executables packaged by Nix are typically patched and relinked with the relevant version of `glibc` libraries from Nix packages (please see the documentation on [patchelf](https://github.com/NixOS/patchelf) for more details). + +### File type groups {#multiple-output-file-type-groups} + +The support code currently recognizes some particular kinds of outputs and either instructs the build system of the package to put files into their desired outputs or it moves the files during the fixup phase. Each group of file types has an `outputFoo` variable specifying the output name where they should go. If that variable isn’t defined by the derivation writer, it is guessed – a default output name is defined, falling back to other possibilities if the output isn’t defined. + +#### `$outputDev` {#outputdev} + +is for development-only files. These include C(++) headers (`include/`), pkg-config (`lib/pkgconfig/`), cmake (`lib/cmake/`) and aclocal files (`share/aclocal/`). They go to `dev` or `out` by default. + +#### `$outputBin` {#outputbin} + +is meant for user-facing binaries, typically residing in `bin/`. They go to `bin` or `out` by default. + +#### `$outputLib` {#outputlib} + +is meant for libraries, typically residing in `lib/` and `libexec/`. They go to `lib` or `out` by default. + +#### `$outputDoc` {#outputdoc} + +is for user documentation, typically residing in `share/doc/`. It goes to `doc` or `out` by default. + +#### `$outputDevdoc` {#outputdevdoc} + +is for _developer_ documentation. Currently we count gtk-doc and devhelp books, typically residing in `share/gtk-doc/` and `share/devhelp/`, in there. It goes to `devdoc` or is removed (!) by default. This is because e.g. gtk-doc tends to be rather large and completely unused by nixpkgs users. + +#### `$outputMan` {#outputman} + +is for man pages (except for section 3), typically residing in `share/man/man[0-9]/`. They go to `man` or `$outputBin` by default. + +#### `$outputDevman` {#outputdevman} + +is for section 3 man pages, typically residing in `share/man/man[0-9]/`. They go to `devman` or `$outputMan` by default. + +#### `$outputInfo` {#outputinfo} + +is for info pages, typically residing in `share/info/`. They go to `info` or `$outputBin` by default. + +### Common caveats {#sec-multiple-outputs-caveats} + +- Some configure scripts don’t like some of the parameters passed by default by the framework, e.g. `--docdir=/foo/bar`. You can disable this by setting `setOutputFlags = false;`. + +- The outputs of a single derivation can retain references to each other, but note that circular references are not allowed. (And each strongly-connected component would act as a single output anyway.) + +- Most of split packages contain their core functionality in libraries. These libraries tend to refer to various kind of data that typically gets into `out`, e.g. locale strings, so there is often no advantage in separating the libraries into `lib`, as keeping them in `out` is easier. + +- Some packages have hidden assumptions on install paths, which complicates splitting. diff --git a/nixpkgs/doc/stdenv/platform-notes.chapter.md b/nixpkgs/doc/stdenv/platform-notes.chapter.md new file mode 100644 index 000000000000..b47f5af349b8 --- /dev/null +++ b/nixpkgs/doc/stdenv/platform-notes.chapter.md @@ -0,0 +1,67 @@ +# Platform Notes {#chap-platform-notes} + +## Darwin (macOS) {#sec-darwin} + +Some common issues when packaging software for Darwin: + +- The Darwin `stdenv` uses clang instead of gcc. When referring to the compiler `$CC` or `cc` will work in both cases. Some builds hardcode gcc/g++ in their build scripts, that can usually be fixed with using something like `makeFlags = [ "CC=cc" ];` or by patching the build scripts. + + ```nix + stdenv.mkDerivation { + name = "libfoo-1.2.3"; + # ... + buildPhase = '' + $CC -o hello hello.c + ''; + } + ``` + +- On Darwin, libraries are linked using absolute paths, libraries are resolved by their `install_name` at link time. Sometimes packages won’t set this correctly causing the library lookups to fail at runtime. This can be fixed by adding extra linker flags or by running `install_name_tool -id` during the `fixupPhase`. + + ```nix + stdenv.mkDerivation { + name = "libfoo-1.2.3"; + # ... + makeFlags = lib.optional stdenv.isDarwin "LDFLAGS=-Wl,-install_name,$(out)/lib/libfoo.dylib"; + } + ``` + +- Even if the libraries are linked using absolute paths and resolved via their `install_name` correctly, tests can sometimes fail to run binaries. This happens because the `checkPhase` runs before the libraries are installed. + + This can usually be solved by running the tests after the `installPhase` or alternatively by using `DYLD_LIBRARY_PATH`. More information about this variable can be found in the *dyld(1)* manpage. + + ``` + dyld: Library not loaded: /nix/store/7hnmbscpayxzxrixrgxvvlifzlxdsdir-jq-1.5-lib/lib/libjq.1.dylib + Referenced from: /private/tmp/nix-build-jq-1.5.drv-0/jq-1.5/tests/../jq + Reason: image not found + ./tests/jqtest: line 5: 75779 Abort trap: 6 + ``` + + ```nix + stdenv.mkDerivation { + name = "libfoo-1.2.3"; + # ... + doInstallCheck = true; + installCheckTarget = "check"; + } + ``` + +- Some packages assume xcode is available and use `xcrun` to resolve build tools like `clang`, etc. This causes errors like `xcode-select: error: no developer tools were found at '/Applications/Xcode.app'` while the build doesn’t actually depend on xcode. + + ```nix + stdenv.mkDerivation { + name = "libfoo-1.2.3"; + # ... + prePatch = '' + substituteInPlace Makefile \ + --replace '/usr/bin/xcrun clang' clang + ''; + } + ``` + + The package `xcbuild` can be used to build projects that really depend on Xcode. However, this replacement is not 100% compatible with Xcode and can occasionally cause issues. + +- x86_64-darwin uses the 10.12 SDK by default, but some software is not compatible with that version of the SDK. In that case, + the 11.0 SDK used by aarch64-darwin is available for use on x86_64-darwin. To use it, reference `apple_sdk_11_0` instead of + `apple_sdk` in your derivation and use `pkgs.darwin.apple_sdk_11_0.callPackage` instead of `pkgs.callPackage`. On Linux, this will + have the same effect as `pkgs.callPackage`, so you can use `pkgs.darwin.apple_sdk_11_0.callPackage` regardless of platform. diff --git a/nixpkgs/doc/stdenv/stdenv.chapter.md b/nixpkgs/doc/stdenv/stdenv.chapter.md new file mode 100644 index 000000000000..0b3776b530ca --- /dev/null +++ b/nixpkgs/doc/stdenv/stdenv.chapter.md @@ -0,0 +1,1475 @@ +# The Standard Environment {#chap-stdenv} + +The standard build environment in the Nix Packages collection provides an environment for building Unix packages that does a lot of common build tasks automatically. In fact, for Unix packages that use the standard `./configure; make; make install` build interface, you don’t need to write a build script at all; the standard environment does everything automatically. If `stdenv` doesn’t do what you need automatically, you can easily customise or override the various build phases. + +## Using `stdenv` {#sec-using-stdenv} + +To build a package with the standard environment, you use the function `stdenv.mkDerivation`, instead of the primitive built-in function `derivation`, e.g. + +```nix +stdenv.mkDerivation { + name = "libfoo-1.2.3"; + src = fetchurl { + url = "http://example.org/libfoo-1.2.3.tar.bz2"; + hash = "sha256-tWxU/LANbQE32my+9AXyt3nCT7NBVfJ45CX757EMT3Q="; + }; +} +``` + +(`stdenv` needs to be in scope, so if you write this in a separate Nix expression from `pkgs/all-packages.nix`, you need to pass it as a function argument.) Specifying a `name` and a `src` is the absolute minimum Nix requires. For convenience, you can also use `pname` and `version` attributes and `mkDerivation` will automatically set `name` to `"${pname}-${version}"` by default. +**Since [RFC 0035](https://github.com/NixOS/rfcs/pull/35), this is preferred for packages in Nixpkgs**, as it allows us to reuse the version easily: + +```nix +stdenv.mkDerivation rec { + pname = "libfoo"; + version = "1.2.3"; + src = fetchurl { + url = "http://example.org/libfoo-source-${version}.tar.bz2"; + hash = "sha256-tWxU/LANbQE32my+9AXyt3nCT7NBVfJ45CX757EMT3Q="; + }; +} +``` + +Many packages have dependencies that are not provided in the standard environment. It’s usually sufficient to specify those dependencies in the `buildInputs` attribute: + +```nix +stdenv.mkDerivation { + pname = "libfoo"; + version = "1.2.3"; + ... + buildInputs = [libbar perl ncurses]; +} +``` + +This attribute ensures that the `bin` subdirectories of these packages appear in the `PATH` environment variable during the build, that their `include` subdirectories are searched by the C compiler, and so on. (See [](#ssec-setup-hooks) for details.) + +Often it is necessary to override or modify some aspect of the build. To make this easier, the standard environment breaks the package build into a number of *phases*, all of which can be overridden or modified individually: unpacking the sources, applying patches, configuring, building, and installing. (There are some others; see [](#sec-stdenv-phases).) For instance, a package that doesn’t supply a makefile but instead has to be compiled "manually" could be handled like this: + +```nix +stdenv.mkDerivation { + pname = "fnord"; + version = "4.5"; + ... + buildPhase = '' + gcc foo.c -o foo + ''; + installPhase = '' + mkdir -p $out/bin + cp foo $out/bin + ''; +} +``` + +(Note the use of `''`-style string literals, which are very convenient for large multi-line script fragments because they don’t need escaping of `"` and `\`, and because indentation is intelligently removed.) + +There are many other attributes to customise the build. These are listed in [](#ssec-stdenv-attributes). + +While the standard environment provides a generic builder, you can still supply your own build script: + +```nix +stdenv.mkDerivation { + pname = "libfoo"; + version = "1.2.3"; + ... + builder = ./builder.sh; +} +``` + +where the builder can do anything it wants, but typically starts with + +```bash +source $stdenv/setup +``` + +to let `stdenv` set up the environment (e.g. by resetting `PATH` and populating it from build inputs). If you want, you can still use `stdenv`’s generic builder: + +```bash +source $stdenv/setup + +buildPhase() { + echo "... this is my custom build phase ..." + gcc foo.c -o foo +} + +installPhase() { + mkdir -p $out/bin + cp foo $out/bin +} + +genericBuild +``` + +### Building a `stdenv` package in `nix-shell` {#sec-building-stdenv-package-in-nix-shell} + +To build a `stdenv` package in a [`nix-shell`](https://nixos.org/manual/nix/unstable/command-ref/nix-shell.html), use + +```bash +nix-shell '<nixpkgs>' -A some_package +eval "${unpackPhase:-unpackPhase}" +cd $sourceRoot +eval "${patchPhase:-patchPhase}" +eval "${configurePhase:-configurePhase}" +eval "${buildPhase:-buildPhase}" +``` + +To modify a [phase](#sec-stdenv-phases), first print it with + +```bash +type buildPhase +``` + +then change it in a text editor, and paste it back to the terminal. + +## Tools provided by `stdenv` {#sec-tools-of-stdenv} + +The standard environment provides the following packages: + +- The GNU C Compiler, configured with C and C++ support. +- GNU coreutils (contains a few dozen standard Unix commands). +- GNU findutils (contains `find`). +- GNU diffutils (contains `diff`, `cmp`). +- GNU `sed`. +- GNU `grep`. +- GNU `awk`. +- GNU `tar`. +- `gzip`, `bzip2` and `xz`. +- GNU Make. +- Bash. This is the shell used for all builders in the Nix Packages collection. Not using `/bin/sh` removes a large source of portability problems. +- The `patch` command. + +On Linux, `stdenv` also includes the `patchelf` utility. + +## Specifying dependencies {#ssec-stdenv-dependencies} + +Build systems often require more dependencies than just what `stdenv` provides. This section describes attributes accepted by `stdenv.mkDerivation` that can be used to make these dependencies available to the build system. + +### Overview {#ssec-stdenv-dependencies-overview} + +A full reference of the different kinds of dependencies is provided in [](#ssec-stdenv-dependencies-reference), but here is an overview of the most common ones. +It should cover most use cases. + +Add dependencies to `nativeBuildInputs` if they are executed during the build: +- those which are needed on `$PATH` during the build, for example `cmake` and `pkg-config` +- [setup hooks](#ssec-setup-hooks), for example [`makeWrapper`](#fun-makeWrapper) +- interpreters needed by [`patchShebangs`](#patch-shebangs.sh) for build scripts (with the `--build` flag), which can be the case for e.g. `perl` + +Add dependencies to `buildInputs` if they will end up copied or linked into the final output or otherwise used at runtime: +- libraries used by compilers, for example `zlib`, +- interpreters needed by [`patchShebangs`](#patch-shebangs.sh) for scripts which are installed, which can be the case for e.g. `perl` + +::: {.note} +These criteria are independent. + +For example, software using Wayland usually needs the `wayland` library at runtime, so `wayland` should be added to `buildInputs`. +But it also executes the `wayland-scanner` program as part of the build to generate code, so `wayland` should also be added to `nativeBuildInputs`. +::: + +Dependencies needed only to run tests are similarly classified between native (executed during build) and non-native (executed at runtime): +- `nativeCheckInputs` for test tools needed on `$PATH` (such as `ctest`) and [setup hooks](#ssec-setup-hooks) (for example [`pytestCheckHook`](#python)) +- `checkInputs` for libraries linked into test executables (for example the `qcheck` OCaml package) + +These dependencies are only injected when [`doCheck`](#var-stdenv-doCheck) is set to `true`. + +#### Example {#ssec-stdenv-dependencies-overview-example} + +Consider for example this simplified derivation for `solo5`, a sandboxing tool: +```nix +stdenv.mkDerivation rec { + pname = "solo5"; + version = "0.7.5"; + + src = fetchurl { + url = "https://github.com/Solo5/solo5/releases/download/v${version}/solo5-v${version}.tar.gz"; + sha256 = "sha256-viwrS9lnaU8sTGuzK/+L/PlMM/xRRtgVuK5pixVeDEw="; + }; + + nativeBuildInputs = [ makeWrapper pkg-config ]; + buildInputs = [ libseccomp ]; + + postInstall = '' + substituteInPlace $out/bin/solo5-virtio-mkimage \ + --replace "/usr/lib/syslinux" "${syslinux}/share/syslinux" \ + --replace "/usr/share/syslinux" "${syslinux}/share/syslinux" \ + --replace "cp " "cp --no-preserve=mode " + + wrapProgram $out/bin/solo5-virtio-mkimage \ + --prefix PATH : ${lib.makeBinPath [ dosfstools mtools parted syslinux ]} + ''; + + doCheck = true; + nativeCheckInputs = [ util-linux qemu ]; + checkPhase = '' [elided] ''; +} +``` + +- `makeWrapper` is a setup hook, i.e., a shell script sourced by the generic builder of `stdenv`. + It is thus executed during the build and must be added to `nativeBuildInputs`. +- `pkg-config` is a build tool which the configure script of `solo5` expects to be on `$PATH` during the build: + therefore, it must be added to `nativeBuildInputs`. +- `libseccomp` is a library linked into `$out/bin/solo5-elftool`. + As it is used at runtime, it must be added to `buildInputs`. +- Tests need `qemu` and `getopt` (from `util-linux`) on `$PATH`, these must be added to `nativeCheckInputs`. +- Some dependencies are injected directly in the shell code of phases: `syslinux`, `dosfstools`, `mtools`, and `parted`. +In this specific case, they will end up in the output of the derivation (`$out` here). +As Nix marks dependencies whose absolute path is present in the output as runtime dependencies, adding them to `buildInputs` is not required. + +For more complex cases, like libraries linked into an executable which is then executed as part of the build system, see [](#ssec-stdenv-dependencies-reference). + +### Reference {#ssec-stdenv-dependencies-reference} + +As described in the Nix manual, almost any `*.drv` store path in a derivation’s attribute set will induce a dependency on that derivation. `mkDerivation`, however, takes a few attributes intended to include all the dependencies of a package. This is done both for structure and consistency, but also so that certain other setup can take place. For example, certain dependencies need their bin directories added to the `PATH`. That is built-in, but other setup is done via a pluggable mechanism that works in conjunction with these dependency attributes. See [](#ssec-setup-hooks) for details. + +Dependencies can be broken down along three axes: their host and target platforms relative to the new derivation’s, and whether they are propagated. The platform distinctions are motivated by cross compilation; see [](#chap-cross) for exactly what each platform means. [^footnote-stdenv-ignored-build-platform] But even if one is not cross compiling, the platforms imply whether or not the dependency is needed at run-time or build-time, a concept that makes perfect sense outside of cross compilation. By default, the run-time/build-time distinction is just a hint for mental clarity, but with `strictDeps` set it is mostly enforced even in the native case. + +The extension of `PATH` with dependencies, alluded to above, proceeds according to the relative platforms alone. The process is carried out only for dependencies whose host platform matches the new derivation’s build platform i.e. dependencies which run on the platform where the new derivation will be built. [^footnote-stdenv-native-dependencies-in-path] For each dependency \<dep\> of those dependencies, `dep/bin`, if present, is added to the `PATH` environment variable. + +A dependency is said to be **propagated** when some of its other-transitive (non-immediate) downstream dependencies also need it as an immediate dependency. +[^footnote-stdenv-propagated-dependencies] + +It is important to note that dependencies are not necessarily propagated as the same sort of dependency that they were before, but rather as the corresponding sort so that the platform rules still line up. To determine the exact rules for dependency propagation, we start by assigning to each dependency a couple of ternary numbers (`-1` for `build`, `0` for `host`, and `1` for `target`) representing its [dependency type](#possible-dependency-types), which captures how its host and target platforms are each "offset" from the depending derivation’s host and target platforms. The following table summarize the different combinations that can be obtained: + +| `host → target` | attribute name | offset | +| ------------------- | ------------------- | -------- | +| `build --> build` | `depsBuildBuild` | `-1, -1` | +| `build --> host` | `nativeBuildInputs` | `-1, 0` | +| `build --> target` | `depsBuildTarget` | `-1, 1` | +| `host --> host` | `depsHostHost` | `0, 0` | +| `host --> target` | `buildInputs` | `0, 1` | +| `target --> target` | `depsTargetTarget` | `1, 1` | + +Algorithmically, we traverse propagated inputs, accumulating every propagated dependency’s propagated dependencies and adjusting them to account for the “shift in perspective” described by the current dependency’s platform offsets. This results is sort of a transitive closure of the dependency relation, with the offsets being approximately summed when two dependency links are combined. We also prune transitive dependencies whose combined offsets go out-of-bounds, which can be viewed as a filter over that transitive closure removing dependencies that are blatantly absurd. + +We can define the process precisely with [Natural Deduction](https://en.wikipedia.org/wiki/Natural_deduction) using the inference rules. This probably seems a bit obtuse, but so is the bash code that actually implements it! [^footnote-stdenv-find-inputs-location] They’re confusing in very different ways so… hopefully if something doesn’t make sense in one presentation, it will in the other! + +``` +let mapOffset(h, t, i) = i + (if i <= 0 then h else t - 1) + +propagated-dep(h0, t0, A, B) +propagated-dep(h1, t1, B, C) +h0 + h1 in {-1, 0, 1} +h0 + t1 in {-1, 0, 1} +-------------------------------------- Transitive property +propagated-dep(mapOffset(h0, t0, h1), + mapOffset(h0, t0, t1), + A, C) +``` + +``` +let mapOffset(h, t, i) = i + (if i <= 0 then h else t - 1) + +dep(h0, t0, A, B) +propagated-dep(h1, t1, B, C) +h0 + h1 in {-1, 0, 1} +h0 + t1 in {-1, 0, -1} +----------------------------- Take immediate dependencies' propagated dependencies +propagated-dep(mapOffset(h0, t0, h1), + mapOffset(h0, t0, t1), + A, C) +``` + +``` +propagated-dep(h, t, A, B) +----------------------------- Propagated dependencies count as dependencies +dep(h, t, A, B) +``` + +Some explanation of this monstrosity is in order. In the common case, the target offset of a dependency is the successor to the target offset: `t = h + 1`. That means that: + +``` +let f(h, t, i) = i + (if i <= 0 then h else t - 1) +let f(h, h + 1, i) = i + (if i <= 0 then h else (h + 1) - 1) +let f(h, h + 1, i) = i + (if i <= 0 then h else h) +let f(h, h + 1, i) = i + h +``` + +This is where “sum-like” comes in from above: We can just sum all of the host offsets to get the host offset of the transitive dependency. The target offset is the transitive dependency is simply the host offset + 1, just as it was with the dependencies composed to make this transitive one; it can be ignored as it doesn’t add any new information. + +Because of the bounds checks, the uncommon cases are `h = t` and `h + 2 = t`. In the former case, the motivation for `mapOffset` is that since its host and target platforms are the same, no transitive dependency of it should be able to “discover” an offset greater than its reduced target offsets. `mapOffset` effectively “squashes” all its transitive dependencies’ offsets so that none will ever be greater than the target offset of the original `h = t` package. In the other case, `h + 1` is skipped over between the host and target offsets. Instead of squashing the offsets, we need to “rip” them apart so no transitive dependencies’ offset is that one. + +Overall, the unifying theme here is that propagation shouldn’t be introducing transitive dependencies involving platforms the depending package is unaware of. \[One can imagine the depending package asking for dependencies with the platforms it knows about; other platforms it doesn’t know how to ask for. The platform description in that scenario is a kind of unforgeable capability.\] The offset bounds checking and definition of `mapOffset` together ensure that this is the case. Discovering a new offset is discovering a new platform, and since those platforms weren’t in the derivation “spec” of the needing package, they cannot be relevant. From a capability perspective, we can imagine that the host and target platforms of a package are the capabilities a package requires, and the depending package must provide the capability to the dependency. + +#### Variables specifying dependencies {#variables-specifying-dependencies} + +##### `depsBuildBuild` {#var-stdenv-depsBuildBuild} + +A list of dependencies whose host and target platforms are the new derivation’s build platform. These are programs and libraries used at build time that produce programs and libraries also used at build time. If the dependency doesn’t care about the target platform (i.e. isn’t a compiler or similar tool), put it in `nativeBuildInputs` instead. The most common use of this `buildPackages.stdenv.cc`, the default C compiler for this role. That example crops up more than one might think in old commonly used C libraries. + +Since these packages are able to be run at build-time, they are always added to the `PATH`, as described above. But since these packages are only guaranteed to be able to run then, they shouldn’t persist as run-time dependencies. This isn’t currently enforced, but could be in the future. + +##### `nativeBuildInputs` {#var-stdenv-nativeBuildInputs} + +A list of dependencies whose host platform is the new derivation’s build platform, and target platform is the new derivation’s host platform. These are programs and libraries used at build-time that, if they are a compiler or similar tool, produce code to run at run-time—i.e. tools used to build the new derivation. If the dependency doesn’t care about the target platform (i.e. isn’t a compiler or similar tool), put it here, rather than in `depsBuildBuild` or `depsBuildTarget`. This could be called `depsBuildHost` but `nativeBuildInputs` is used for historical continuity. + +Since these packages are able to be run at build-time, they are added to the `PATH`, as described above. But since these packages are only guaranteed to be able to run then, they shouldn’t persist as run-time dependencies. This isn’t currently enforced, but could be in the future. + +##### `depsBuildTarget` {#var-stdenv-depsBuildTarget} + +A list of dependencies whose host platform is the new derivation’s build platform, and target platform is the new derivation’s target platform. These are programs used at build time that produce code to run with code produced by the depending package. Most commonly, these are tools used to build the runtime or standard library that the currently-being-built compiler will inject into any code it compiles. In many cases, the currently-being-built-compiler is itself employed for that task, but when that compiler won’t run (i.e. its build and host platform differ) this is not possible. Other times, the compiler relies on some other tool, like binutils, that is always built separately so that the dependency is unconditional. + +This is a somewhat confusing concept to wrap one’s head around, and for good reason. As the only dependency type where the platform offsets, `-1` and `1`, are not adjacent integers, it requires thinking of a bootstrapping stage *two* away from the current one. It and its use-case go hand in hand and are both considered poor form: try to not need this sort of dependency, and try to avoid building standard libraries and runtimes in the same derivation as the compiler produces code using them. Instead strive to build those like a normal library, using the newly-built compiler just as a normal library would. In short, do not use this attribute unless you are packaging a compiler and are sure it is needed. + +Since these packages are able to run at build time, they are added to the `PATH`, as described above. But since these packages are only guaranteed to be able to run then, they shouldn’t persist as run-time dependencies. This isn’t currently enforced, but could be in the future. + +##### `depsHostHost` {#var-stdenv-depsHostHost} + +A list of dependencies whose host and target platforms match the new derivation’s host platform. In practice, this would usually be tools used by compilers for macros or a metaprogramming system, or libraries used by the macros or metaprogramming code itself. It’s always preferable to use a `depsBuildBuild` dependency in the derivation being built over a `depsHostHost` on the tool doing the building for this purpose. + +##### `buildInputs` {#var-stdenv-buildInputs} + +A list of dependencies whose host platform and target platform match the new derivation’s. This would be called `depsHostTarget` but for historical continuity. If the dependency doesn’t care about the target platform (i.e. isn’t a compiler or similar tool), put it here, rather than in `depsBuildBuild`. + +These are often programs and libraries used by the new derivation at *run*-time, but that isn’t always the case. For example, the machine code in a statically-linked library is only used at run-time, but the derivation containing the library is only needed at build-time. Even in the dynamic case, the library may also be needed at build-time to appease the linker. + +##### `depsTargetTarget` {#var-stdenv-depsTargetTarget} + +A list of dependencies whose host platform matches the new derivation’s target platform. These are packages that run on the target platform, e.g. the standard library or run-time deps of standard library that a compiler insists on knowing about. It’s poor form in almost all cases for a package to depend on another from a future stage \[future stage corresponding to positive offset\]. Do not use this attribute unless you are packaging a compiler and are sure it is needed. + +##### `depsBuildBuildPropagated` {#var-stdenv-depsBuildBuildPropagated} + +The propagated equivalent of `depsBuildBuild`. This perhaps never ought to be used, but it is included for consistency \[see below for the others\]. + +##### `propagatedNativeBuildInputs` {#var-stdenv-propagatedNativeBuildInputs} + +The propagated equivalent of `nativeBuildInputs`. This would be called `depsBuildHostPropagated` but for historical continuity. For example, if package `Y` has `propagatedNativeBuildInputs = [X]`, and package `Z` has `buildInputs = [Y]`, then package `Z` will be built as if it included package `X` in its `nativeBuildInputs`. If instead, package `Z` has `nativeBuildInputs = [Y]`, then `Z` will be built as if it included `X` in the `depsBuildBuild` of package `Z`, because of the sum of the two `-1` host offsets. + +##### `depsBuildTargetPropagated` {#var-stdenv-depsBuildTargetPropagated} + +The propagated equivalent of `depsBuildTarget`. This is prefixed for the same reason of alerting potential users. + +##### `depsHostHostPropagated` {#var-stdenv-depsHostHostPropagated} + +The propagated equivalent of `depsHostHost`. + +##### `propagatedBuildInputs` {#var-stdenv-propagatedBuildInputs} + +The propagated equivalent of `buildInputs`. This would be called `depsHostTargetPropagated` but for historical continuity. + +##### `depsTargetTargetPropagated` {#var-stdenv-depsTargetTargetPropagated} + +The propagated equivalent of `depsTargetTarget`. This is prefixed for the same reason of alerting potential users. + +## Attributes {#ssec-stdenv-attributes} + +### Variables affecting `stdenv` initialisation {#variables-affecting-stdenv-initialisation} + +#### `NIX_DEBUG` {#var-stdenv-NIX_DEBUG} + +A number between 0 and 7 indicating how much information to log. If set to 1 or higher, `stdenv` will print moderate debugging information during the build. In particular, the `gcc` and `ld` wrapper scripts will print out the complete command line passed to the wrapped tools. If set to 6 or higher, the `stdenv` setup script will be run with `set -x` tracing. If set to 7 or higher, the `gcc` and `ld` wrapper scripts will also be run with `set -x` tracing. + +### Attributes affecting build properties {#attributes-affecting-build-properties} + +#### `enableParallelBuilding` {#var-stdenv-enableParallelBuilding} + +If set to `true`, `stdenv` will pass specific flags to `make` and other build tools to enable parallel building with up to `build-cores` workers. + +Unless set to `false`, some build systems with good support for parallel building including `cmake`, `meson`, and `qmake` will set it to `true`. + +### Special variables {#special-variables} + +#### `passthru` {#var-stdenv-passthru} + +This is an attribute set which can be filled with arbitrary values. For example: + +```nix +passthru = { + foo = "bar"; + baz = { + value1 = 4; + value2 = 5; + }; +} +``` + +Values inside it are not passed to the builder, so you can change them without triggering a rebuild. However, they can be accessed outside of a derivation directly, as if they were set inside a derivation itself, e.g. `hello.baz.value1`. We don’t specify any usage or schema of `passthru` - it is meant for values that would be useful outside the derivation in other parts of a Nix expression (e.g. in other derivations). An example would be to convey some specific dependency of your derivation which contains a program with plugins support. Later, others who make derivations with plugins can use passed-through dependency to ensure that their plugin would be binary-compatible with built program. + +#### `passthru.updateScript` {#var-passthru-updateScript} + +A script to be run by `maintainers/scripts/update.nix` when the package is matched. The attribute can contain one of the following: + +- []{#var-passthru-updateScript-command} an executable file, either on the file system: + + ```nix + passthru.updateScript = ./update.sh; + ``` + + or inside the expression itself: + + ```nix + passthru.updateScript = writeScript "update-zoom-us" '' + #!/usr/bin/env nix-shell + #!nix-shell -i bash -p curl pcre common-updater-scripts + + set -eu -o pipefail + + version="$(curl -sI https://zoom.us/client/latest/zoom_x86_64.tar.xz | grep -Fi 'Location:' | pcregrep -o1 '/(([0-9]\.?)+)/')" + update-source-version zoom-us "$version" + ''; + ``` + +- a list, a script followed by arguments to be passed to it: + + ```nix + passthru.updateScript = [ ../../update.sh pname "--requested-release=unstable" ]; + ``` + +- an attribute set containing: + - [`command`]{#var-passthru-updateScript-set-command} – a string or list in the [format expected by `passthru.updateScript`](#var-passthru-updateScript-command). + - [`attrPath`]{#var-passthru-updateScript-set-attrPath} (optional) – a string containing the canonical attribute path for the package. If present, it will be passed to the update script instead of the attribute path on which the package was discovered during Nixpkgs traversal. + - [`supportedFeatures`]{#var-passthru-updateScript-set-supportedFeatures} (optional) – a list of the [extra features](#var-passthru-updateScript-supported-features) the script supports. + + ```nix + passthru.updateScript = { + command = [ ../../update.sh pname ]; + attrPath = pname; + supportedFeatures = [ … ]; + }; + ``` + +##### How update scripts are executed? {#var-passthru-updateScript-execution} + +Update scripts are to be invoked by `maintainers/scripts/update.nix` script. You can run `nix-shell maintainers/scripts/update.nix` in the root of Nixpkgs repository for information on how to use it. `update.nix` offers several modes for selecting packages to update (e.g. select by attribute path, traverse Nixpkgs and filter by maintainer, etc.), and it will execute update scripts for all matched packages that have an `updateScript` attribute. + +Each update script will be passed the following environment variables: + +- [`UPDATE_NIX_NAME`]{#var-passthru-updateScript-env-UPDATE_NIX_NAME} – content of the `name` attribute of the updated package. +- [`UPDATE_NIX_PNAME`]{#var-passthru-updateScript-env-UPDATE_NIX_PNAME} – content of the `pname` attribute of the updated package. +- [`UPDATE_NIX_OLD_VERSION`]{#var-passthru-updateScript-env-UPDATE_NIX_OLD_VERSION} – content of the `version` attribute of the updated package. +- [`UPDATE_NIX_ATTR_PATH`]{#var-passthru-updateScript-env-UPDATE_NIX_ATTR_PATH} – attribute path the `update.nix` discovered the package on (or the [canonical `attrPath`](#var-passthru-updateScript-set-attrPath) when available). Example: `pantheon.elementary-terminal` + +::: {.note} +An update script will be usually run from the root of the Nixpkgs repository but you should not rely on that. Also note that `update.nix` executes update scripts in parallel by default so you should avoid running `git commit` or any other commands that cannot handle that. +::: + +::: {.tip} +While update scripts should not create commits themselves, `maintainers/scripts/update.nix` supports automatically creating commits when running it with `--argstr commit true`. If you need to customize commit message, you can have the update script implement [`commit`](#var-passthru-updateScript-commit) feature. +::: + +##### Supported features {#var-passthru-updateScript-supported-features} +###### `commit` {#var-passthru-updateScript-commit} + +This feature allows update scripts to *ask* `update.nix` to create Git commits. + +When support of this feature is declared, whenever the update script exits with `0` return status, it is expected to print a JSON list containing an object described below for each updated attribute to standard output. + +When `update.nix` is run with `--argstr commit true` arguments, it will create a separate commit for each of the objects. An empty list can be returned when the script did not update any files, for example, when the package is already at the latest version. + +The commit object contains the following values: + +- [`attrPath`]{#var-passthru-updateScript-commit-attrPath} – a string containing attribute path. +- [`oldVersion`]{#var-passthru-updateScript-commit-oldVersion} – a string containing old version. +- [`newVersion`]{#var-passthru-updateScript-commit-newVersion} – a string containing new version. +- [`files`]{#var-passthru-updateScript-commit-files} – a non-empty list of file paths (as strings) to add to the commit. +- [`commitBody`]{#var-passthru-updateScript-commit-commitBody} (optional) – a string with extra content to be appended to the default commit message (useful for adding changelog links). +- [`commitMessage`]{#var-passthru-updateScript-commit-commitMessage} (optional) – a string to use instead of the default commit message. + +If the returned array contains exactly one object (e.g. `[{}]`), all values are optional and will be determined automatically. + +```{=docbook} +<example> +<title>Standard output of an update script using commit feature</title> +``` + +```json +[ + { + "attrPath": "volume_key", + "oldVersion": "0.3.11", + "newVersion": "0.3.12", + "files": [ + "/path/to/nixpkgs/pkgs/development/libraries/volume-key/default.nix" + ] + } +] +``` + +```{=docbook} +</example> +``` + +### Recursive attributes in `mkDerivation` {#mkderivation-recursive-attributes} + +If you pass a function to `mkDerivation`, it will receive as its argument the final arguments, including the overrides when reinvoked via `overrideAttrs`. For example: + +```nix +mkDerivation (finalAttrs: { + pname = "hello"; + withFeature = true; + configureFlags = + lib.optionals finalAttrs.withFeature ["--with-feature"]; +}) +``` + +Note that this does not use the `rec` keyword to reuse `withFeature` in `configureFlags`. +The `rec` keyword works at the syntax level and is unaware of overriding. + +Instead, the definition references `finalAttrs`, allowing users to change `withFeature` +consistently with `overrideAttrs`. + +`finalAttrs` also contains the attribute `finalPackage`, which includes the output paths, etc. + +Let's look at a more elaborate example to understand the differences between +various bindings: + +```nix +# `pkg` is the _original_ definition (for illustration purposes) +let pkg = + mkDerivation (finalAttrs: { + # ... + + # An example attribute + packages = []; + + # `passthru.tests` is a commonly defined attribute. + passthru.tests.simple = f finalAttrs.finalPackage; + + # An example of an attribute containing a function + passthru.appendPackages = packages': + finalAttrs.finalPackage.overrideAttrs (newSelf: super: { + packages = super.packages ++ packages'; + }); + + # For illustration purposes; referenced as + # `(pkg.overrideAttrs(x)).finalAttrs` etc in the text below. + passthru.finalAttrs = finalAttrs; + passthru.original = pkg; + }); +in pkg +``` + +Unlike the `pkg` binding in the above example, the `finalAttrs` parameter always references the final attributes. For instance `(pkg.overrideAttrs(x)).finalAttrs.finalPackage` is identical to `pkg.overrideAttrs(x)`, whereas `(pkg.overrideAttrs(x)).original` is the same as the original `pkg`. + +See also the section about [`passthru.tests`](#var-meta-tests). + +## Phases {#sec-stdenv-phases} + +`stdenv.mkDerivation` sets the Nix [derivation](https://nixos.org/manual/nix/stable/expressions/derivations.html#derivations)'s builder to a script that loads the stdenv `setup.sh` bash library and calls `genericBuild`. Most packaging functions rely on this default builder. + +This generic command invokes a number of *phases*. Package builds are split into phases to make it easier to override specific parts of the build (e.g., unpacking the sources or installing the binaries). + +Each phase can be overridden in its entirety either by setting the environment variable `namePhase` to a string containing some shell commands to be executed, or by redefining the shell function `namePhase`. The former is convenient to override a phase from the derivation, while the latter is convenient from a build script. However, typically one only wants to *add* some commands to a phase, e.g. by defining `postInstall` or `preFixup`, as skipping some of the default actions may have unexpected consequences. The default script for each phase is defined in the file `pkgs/stdenv/generic/setup.sh`. + +When overriding a phase, for example `installPhase`, it is important to start with `runHook preInstall` and end it with `runHook postInstall`, otherwise `preInstall` and `postInstall` will not be run. Even if you don't use them directly, it is good practice to do so anyways for downstream users who would want to add a `postInstall` by overriding your derivation. + +While inside an interactive `nix-shell`, if you wanted to run all phases in the order they would be run in an actual build, you can invoke `genericBuild` yourself. + +### Controlling phases {#ssec-controlling-phases} + +There are a number of variables that control what phases are executed and in what order: + +#### Variables affecting phase control {#variables-affecting-phase-control} + +##### `phases` {#var-stdenv-phases} + +Specifies the phases. You can change the order in which phases are executed, or add new phases, by setting this variable. If it’s not set, the default value is used, which is `$prePhases unpackPhase patchPhase $preConfigurePhases configurePhase $preBuildPhases buildPhase checkPhase $preInstallPhases installPhase fixupPhase installCheckPhase $preDistPhases distPhase $postPhases`. + +It is discouraged to set this variable, as it is easy to miss some important functionality hidden in some of the less obviously needed phases (like `fixupPhase` which patches the shebang of scripts). +Usually, if you just want to add a few phases, it’s more convenient to set one of the variables below (such as `preInstallPhases`). + +##### `prePhases` {#var-stdenv-prePhases} + +Additional phases executed before any of the default phases. + +##### `preConfigurePhases` {#var-stdenv-preConfigurePhases} + +Additional phases executed just before the configure phase. + +##### `preBuildPhases` {#var-stdenv-preBuildPhases} + +Additional phases executed just before the build phase. + +##### `preInstallPhases` {#var-stdenv-preInstallPhases} + +Additional phases executed just before the install phase. + +##### `preFixupPhases` {#var-stdenv-preFixupPhases} + +Additional phases executed just before the fixup phase. + +##### `preDistPhases` {#var-stdenv-preDistPhases} + +Additional phases executed just before the distribution phase. + +##### `postPhases` {#var-stdenv-postPhases} + +Additional phases executed after any of the default phases. + +### The unpack phase {#ssec-unpack-phase} + +The unpack phase is responsible for unpacking the source code of the package. The default implementation of `unpackPhase` unpacks the source files listed in the `src` environment variable to the current directory. It supports the following files by default: + +#### Tar files {#tar-files} + +These can optionally be compressed using `gzip` (`.tar.gz`, `.tgz` or `.tar.Z`), `bzip2` (`.tar.bz2`, `.tbz2` or `.tbz`) or `xz` (`.tar.xz`, `.tar.lzma` or `.txz`). + +#### Zip files {#zip-files} + +Zip files are unpacked using `unzip`. However, `unzip` is not in the standard environment, so you should add it to `nativeBuildInputs` yourself. + +#### Directories in the Nix store {#directories-in-the-nix-store} + +These are simply copied to the current directory. The hash part of the file name is stripped, e.g. `/nix/store/1wydxgby13cz...-my-sources` would be copied to `my-sources`. + +Additional file types can be supported by setting the `unpackCmd` variable (see below). + +#### Variables controlling the unpack phase {#variables-controlling-the-unpack-phase} + +##### `srcs` / `src` {#var-stdenv-src} + +The list of source files or directories to be unpacked or copied. One of these must be set. Note that if you use `srcs`, you should also set `sourceRoot` or `setSourceRoot`. + +##### `sourceRoot` {#var-stdenv-sourceRoot} + +After running `unpackPhase`, the generic builder changes the current directory to the directory created by unpacking the sources. If there are multiple source directories, you should set `sourceRoot` to the name of the intended directory. Set `sourceRoot = ".";` if you use `srcs` and control the unpack phase yourself. + +By default the `sourceRoot` is set to `"source"`. If you want to point to a sub-directory inside your project, you therefore need to set `sourceRoot = "source/my-sub-directory"`. + +##### `setSourceRoot` {#var-stdenv-setSourceRoot} + +Alternatively to setting `sourceRoot`, you can set `setSourceRoot` to a shell command to be evaluated by the unpack phase after the sources have been unpacked. This command must set `sourceRoot`. + +##### `preUnpack` {#var-stdenv-preUnpack} + +Hook executed at the start of the unpack phase. + +##### `postUnpack` {#var-stdenv-postUnpack} + +Hook executed at the end of the unpack phase. + +##### `dontUnpack` {#var-stdenv-dontUnpack} + +Set to true to skip the unpack phase. + +##### `dontMakeSourcesWritable` {#var-stdenv-dontMakeSourcesWritable} + +If set to `1`, the unpacked sources are *not* made writable. By default, they are made writable to prevent problems with read-only sources. For example, copied store directories would be read-only without this. + +##### `unpackCmd` {#var-stdenv-unpackCmd} + +The unpack phase evaluates the string `$unpackCmd` for any unrecognised file. The path to the current source file is contained in the `curSrc` variable. + +### The patch phase {#ssec-patch-phase} + +The patch phase applies the list of patches defined in the `patches` variable. + +#### Variables controlling the patch phase {#variables-controlling-the-patch-phase} + +##### `dontPatch` {#var-stdenv-dontPatch} + +Set to true to skip the patch phase. + +##### `patches` {#var-stdenv-patches} + +The list of patches. They must be in the format accepted by the `patch` command, and may optionally be compressed using `gzip` (`.gz`), `bzip2` (`.bz2`) or `xz` (`.xz`). + +##### `patchFlags` {#var-stdenv-patchFlags} + +Flags to be passed to `patch`. If not set, the argument `-p1` is used, which causes the leading directory component to be stripped from the file names in each patch. + +##### `prePatch` {#var-stdenv-prePatch} + +Hook executed at the start of the patch phase. + +##### `postPatch` {#var-stdenv-postPatch} + +Hook executed at the end of the patch phase. + +### The configure phase {#ssec-configure-phase} + +The configure phase prepares the source tree for building. The default `configurePhase` runs `./configure` (typically an Autoconf-generated script) if it exists. + +#### Variables controlling the configure phase {#variables-controlling-the-configure-phase} + +##### `configureScript` {#var-stdenv-configureScript} + +The name of the configure script. It defaults to `./configure` if it exists; otherwise, the configure phase is skipped. This can actually be a command (like `perl ./Configure.pl`). + +##### `configureFlags` {#var-stdenv-configureFlags} + +A list of strings passed as additional arguments to the configure script. + +##### `dontConfigure` {#var-stdenv-dontConfigure} + +Set to true to skip the configure phase. + +##### `configureFlagsArray` {#var-stdenv-configureFlagsArray} + +A shell array containing additional arguments passed to the configure script. You must use this instead of `configureFlags` if the arguments contain spaces. + +##### `dontAddPrefix` {#var-stdenv-dontAddPrefix} + +By default, the flag `--prefix=$prefix` is added to the configure flags. If this is undesirable, set this variable to true. + +##### `prefix` {#var-stdenv-prefix} + +The prefix under which the package must be installed, passed via the `--prefix` option to the configure script. It defaults to `$out`. + +##### `prefixKey` {#var-stdenv-prefixKey} + +The key to use when specifying the prefix. By default, this is set to `--prefix=` as that is used by the majority of packages. + +##### `dontAddStaticConfigureFlags` {#var-stdenv-dontAddStaticConfigureFlags} + +By default, when building statically, stdenv will try to add build system appropriate configure flags to try to enable static builds. + +If this is undesirable, set this variable to true. + +##### `dontAddDisableDepTrack` {#var-stdenv-dontAddDisableDepTrack} + +By default, the flag `--disable-dependency-tracking` is added to the configure flags to speed up Automake-based builds. If this is undesirable, set this variable to true. + +##### `dontFixLibtool` {#var-stdenv-dontFixLibtool} + +By default, the configure phase applies some special hackery to all files called `ltmain.sh` before running the configure script in order to improve the purity of Libtool-based packages [^footnote-stdenv-sys-lib-search-path] . If this is undesirable, set this variable to true. + +##### `dontDisableStatic` {#var-stdenv-dontDisableStatic} + +By default, when the configure script has `--enable-static`, the option `--disable-static` is added to the configure flags. + +If this is undesirable, set this variable to true. It is automatically set to true when building statically, for example through `pkgsStatic`. + +##### `configurePlatforms` {#var-stdenv-configurePlatforms} + +By default, when cross compiling, the configure script has `--build=...` and `--host=...` passed. Packages can instead pass `[ "build" "host" "target" ]` or a subset to control exactly which platform flags are passed. Compilers and other tools can use this to also pass the target platform. [^footnote-stdenv-build-time-guessing-impurity] + +##### `preConfigure` {#var-stdenv-preConfigure} + +Hook executed at the start of the configure phase. + +##### `postConfigure` {#var-stdenv-postConfigure} + +Hook executed at the end of the configure phase. + +### The build phase {#build-phase} + +The build phase is responsible for actually building the package (e.g. compiling it). The default `buildPhase` simply calls `make` if a file named `Makefile`, `makefile` or `GNUmakefile` exists in the current directory (or the `makefile` is explicitly set); otherwise it does nothing. + +#### Variables controlling the build phase {#variables-controlling-the-build-phase} + +##### `dontBuild` {#var-stdenv-dontBuild} + +Set to true to skip the build phase. + +##### `makefile` {#var-stdenv-makefile} + +The file name of the Makefile. + +##### `makeFlags` {#var-stdenv-makeFlags} + +A list of strings passed as additional flags to `make`. These flags are also used by the default install and check phase. For setting make flags specific to the build phase, use `buildFlags` (see below). + +```nix +makeFlags = [ "PREFIX=$(out)" ]; +``` + +::: {.note} +The flags are quoted in bash, but environment variables can be specified by using the make syntax. +::: + +##### `makeFlagsArray` {#var-stdenv-makeFlagsArray} + +A shell array containing additional arguments passed to `make`. You must use this instead of `makeFlags` if the arguments contain spaces, e.g. + +```nix +preBuild = '' + makeFlagsArray+=(CFLAGS="-O0 -g" LDFLAGS="-lfoo -lbar") +''; +``` + +Note that shell arrays cannot be passed through environment variables, so you cannot set `makeFlagsArray` in a derivation attribute (because those are passed through environment variables): you have to define them in shell code. + +##### `buildFlags` / `buildFlagsArray` {#var-stdenv-buildFlags} + +A list of strings passed as additional flags to `make`. Like `makeFlags` and `makeFlagsArray`, but only used by the build phase. + +##### `preBuild` {#var-stdenv-preBuild} + +Hook executed at the start of the build phase. + +##### `postBuild` {#var-stdenv-postBuild} + +Hook executed at the end of the build phase. + +You can set flags for `make` through the `makeFlags` variable. + +Before and after running `make`, the hooks `preBuild` and `postBuild` are called, respectively. + +### The check phase {#ssec-check-phase} + +The check phase checks whether the package was built correctly by running its test suite. The default `checkPhase` calls `make $checkTarget`, but only if the [`doCheck` variable](#var-stdenv-doCheck) is enabled. + +#### Variables controlling the check phase {#variables-controlling-the-check-phase} + +##### `doCheck` {#var-stdenv-doCheck} + +Controls whether the check phase is executed. By default it is skipped, but if `doCheck` is set to true, the check phase is usually executed. Thus you should set + +```nix +doCheck = true; +``` + +in the derivation to enable checks. The exception is cross compilation. Cross compiled builds never run tests, no matter how `doCheck` is set, as the newly-built program won’t run on the platform used to build it. + +##### `makeFlags` / `makeFlagsArray` / `makefile` {#makeflags-makeflagsarray-makefile} + +See the [build phase](#var-stdenv-makeFlags) for details. + +##### `checkTarget` {#var-stdenv-checkTarget} + +The `make` target that runs the tests. +If unset, use `check` if it exists, otherwise `test`; if neither is found, do nothing. + +##### `checkFlags` / `checkFlagsArray` {#var-stdenv-checkFlags} + +A list of strings passed as additional flags to `make`. Like `makeFlags` and `makeFlagsArray`, but only used by the check phase. + +##### `checkInputs` {#var-stdenv-checkInputs} + +A list of host dependencies used by the phase, usually libraries linked into executables built during tests. This gets included in `buildInputs` when `doCheck` is set. + +##### `nativeCheckInputs` {#var-stdenv-nativeCheckInputs} + +A list of native dependencies used by the phase, notably tools needed on `$PATH`. This gets included in `nativeBuildInputs` when `doCheck` is set. + +##### `preCheck` {#var-stdenv-preCheck} + +Hook executed at the start of the check phase. + +##### `postCheck` {#var-stdenv-postCheck} + +Hook executed at the end of the check phase. + +### The install phase {#ssec-install-phase} + +The install phase is responsible for installing the package in the Nix store under `out`. The default `installPhase` creates the directory `$out` and calls `make install`. + +#### Variables controlling the install phase {#variables-controlling-the-install-phase} + +##### `dontInstall` {#var-stdenv-dontInstall} + +Set to true to skip the install phase. + +##### `makeFlags` / `makeFlagsArray` / `makefile` {#makeflags-makeflagsarray-makefile-1} + +See the [build phase](#var-stdenv-makeFlags) for details. + +##### `installTargets` {#var-stdenv-installTargets} + +The make targets that perform the installation. Defaults to `install`. Example: + +```nix +installTargets = "install-bin install-doc"; +``` + +##### `installFlags` / `installFlagsArray` {#var-stdenv-installFlags} + +A list of strings passed as additional flags to `make`. Like `makeFlags` and `makeFlagsArray`, but only used by the install phase. + +##### `preInstall` {#var-stdenv-preInstall} + +Hook executed at the start of the install phase. + +##### `postInstall` {#var-stdenv-postInstall} + +Hook executed at the end of the install phase. + +### The fixup phase {#ssec-fixup-phase} + +The fixup phase performs (Nix-specific) post-processing actions on the files installed under `$out` by the install phase. The default `fixupPhase` does the following: + +- It moves the `man/`, `doc/` and `info/` subdirectories of `$out` to `share/`. +- It strips libraries and executables of debug information. +- On Linux, it applies the `patchelf` command to ELF executables and libraries to remove unused directories from the `RPATH` in order to prevent unnecessary runtime dependencies. +- It rewrites the interpreter paths of shell scripts to paths found in `PATH`. E.g., `/usr/bin/perl` will be rewritten to `/nix/store/some-perl/bin/perl` found in `PATH`. See [](#patch-shebangs.sh) for details. + +#### Variables controlling the fixup phase {#variables-controlling-the-fixup-phase} + +##### `dontFixup` {#var-stdenv-dontFixup} + +Set to true to skip the fixup phase. + +##### `dontStrip` {#var-stdenv-dontStrip} + +If set, libraries and executables are not stripped. By default, they are. + +##### `dontStripHost` {#var-stdenv-dontStripHost} + +Like `dontStrip`, but only affects the `strip` command targeting the package’s host platform. Useful when supporting cross compilation, but otherwise feel free to ignore. + +##### `dontStripTarget` {#var-stdenv-dontStripTarget} + +Like `dontStrip`, but only affects the `strip` command targeting the packages’ target platform. Useful when supporting cross compilation, but otherwise feel free to ignore. + +##### `dontMoveSbin` {#var-stdenv-dontMoveSbin} + +If set, files in `$out/sbin` are not moved to `$out/bin`. By default, they are. + +##### `stripAllList` {#var-stdenv-stripAllList} + +List of directories to search for libraries and executables from which *all* symbols should be stripped. By default, it’s empty. Stripping all symbols is risky, since it may remove not just debug symbols but also ELF information necessary for normal execution. + +##### `stripAllListTarget` {#var-stdenv-stripAllListTarget} + +Like `stripAllList`, but only applies to packages’ target platform. By default, it’s empty. Useful when supporting cross compilation. + +##### `stripAllFlags` {#var-stdenv-stripAllFlags} + +Flags passed to the `strip` command applied to the files in the directories listed in `stripAllList`. Defaults to `-s` (i.e. `--strip-all`). + +##### `stripDebugList` {#var-stdenv-stripDebugList} + +List of directories to search for libraries and executables from which only debugging-related symbols should be stripped. It defaults to `lib lib32 lib64 libexec bin sbin`. + +##### `stripDebugListTarget` {#var-stdenv-stripDebugListTarget} + +Like `stripDebugList`, but only applies to packages’ target platform. By default, it’s empty. Useful when supporting cross compilation. + +##### `stripDebugFlags` {#var-stdenv-stripDebugFlags} + +Flags passed to the `strip` command applied to the files in the directories listed in `stripDebugList`. Defaults to `-S` (i.e. `--strip-debug`). + +##### `dontPatchELF` {#var-stdenv-dontPatchELF} + +If set, the `patchelf` command is not used to remove unnecessary `RPATH` entries. Only applies to Linux. + +##### `dontPatchShebangs` {#var-stdenv-dontPatchShebangs} + +If set, scripts starting with `#!` do not have their interpreter paths rewritten to paths in the Nix store. See [](#patch-shebangs.sh) on how patching shebangs works. + +##### `dontPruneLibtoolFiles` {#var-stdenv-dontPruneLibtoolFiles} + +If set, libtool `.la` files associated with shared libraries won’t have their `dependency_libs` field cleared. + +##### `forceShare` {#var-stdenv-forceShare} + +The list of directories that must be moved from `$out` to `$out/share`. Defaults to `man doc info`. + +##### `setupHook` {#var-stdenv-setupHook} + +A package can export a [setup hook](#ssec-setup-hooks) by setting this variable. The setup hook, if defined, is copied to `$out/nix-support/setup-hook`. Environment variables are then substituted in it using `substituteAll`. + +##### `preFixup` {#var-stdenv-preFixup} + +Hook executed at the start of the fixup phase. + +##### `postFixup` {#var-stdenv-postFixup} + +Hook executed at the end of the fixup phase. + +##### `separateDebugInfo` {#stdenv-separateDebugInfo} + +If set to `true`, the standard environment will enable debug information in C/C++ builds. After installation, the debug information will be separated from the executables and stored in the output named `debug`. (This output is enabled automatically; you don’t need to set the `outputs` attribute explicitly.) To be precise, the debug information is stored in `debug/lib/debug/.build-id/XX/YYYY…`, where \<XXYYYY…\> is the \<build ID\> of the binary — a SHA-1 hash of the contents of the binary. Debuggers like GDB use the build ID to look up the separated debug information. + +For example, with GDB, you can add + +``` +set debug-file-directory ~/.nix-profile/lib/debug +``` + +to `~/.gdbinit`. GDB will then be able to find debug information installed via `nix-env -i`. + +### The installCheck phase {#ssec-installCheck-phase} + +The installCheck phase checks whether the package was installed correctly by running its test suite against the installed directories. The default `installCheck` calls `make installcheck`. + +It is often better to add tests that are not part of the source distribution to `passthru.tests` (see <xref linkend="var-meta-tests"/>). This avoids adding overhead to every build and enables us to run them independently. + +#### Variables controlling the installCheck phase {#variables-controlling-the-installcheck-phase} + +##### `doInstallCheck` {#var-stdenv-doInstallCheck} + +Controls whether the installCheck phase is executed. By default it is skipped, but if `doInstallCheck` is set to true, the installCheck phase is usually executed. Thus you should set + +```nix +doInstallCheck = true; +``` + +in the derivation to enable install checks. The exception is cross compilation. Cross compiled builds never run tests, no matter how `doInstallCheck` is set, as the newly-built program won’t run on the platform used to build it. + +##### `installCheckTarget` {#var-stdenv-installCheckTarget} + +The make target that runs the install tests. Defaults to `installcheck`. + +##### `installCheckFlags` / `installCheckFlagsArray` {#var-stdenv-installCheckFlags} + +A list of strings passed as additional flags to `make`. Like `makeFlags` and `makeFlagsArray`, but only used by the installCheck phase. + +##### `installCheckInputs` {#var-stdenv-installCheckInputs} + +A list of host dependencies used by the phase, usually libraries linked into executables built during tests. This gets included in `buildInputs` when `doInstallCheck` is set. + +##### `nativeInstallCheckInputs` {#var-stdenv-nativeInstallCheckInputs} + +A list of native dependencies used by the phase, notably tools needed on `$PATH`. This gets included in `nativeBuildInputs` when `doInstallCheck` is set. + +##### `preInstallCheck` {#var-stdenv-preInstallCheck} + +Hook executed at the start of the installCheck phase. + +##### `postInstallCheck` {#var-stdenv-postInstallCheck} + +Hook executed at the end of the installCheck phase. + +### The distribution phase {#ssec-distribution-phase} + +The distribution phase is intended to produce a source distribution of the package. The default `distPhase` first calls `make dist`, then it copies the resulting source tarballs to `$out/tarballs/`. This phase is only executed if the attribute `doDist` is set. + +#### Variables controlling the distribution phase {#variables-controlling-the-distribution-phase} + +##### `doDist` {#var-stdenv-doDist} + +If set, the distribution phase is executed. + +##### `distTarget` {#var-stdenv-distTarget} + +The make target that produces the distribution. Defaults to `dist`. + +##### `distFlags` / `distFlagsArray` {#var-stdenv-distFlags} + +Additional flags passed to `make`. + +##### `tarballs` {#var-stdenv-tarballs} + +The names of the source distribution files to be copied to `$out/tarballs/`. It can contain shell wildcards. The default is `*.tar.gz`. + +##### `dontCopyDist` {#var-stdenv-dontCopyDist} + +If set, no files are copied to `$out/tarballs/`. + +##### `preDist` {#var-stdenv-preDist} + +Hook executed at the start of the distribution phase. + +##### `postDist` {#var-stdenv-postDist} + +Hook executed at the end of the distribution phase. + +## Shell functions and utilities {#ssec-stdenv-functions} + +The standard environment provides a number of useful functions. + +### `makeWrapper` \<executable\> \<wrapperfile\> \<args\> {#fun-makeWrapper} + +Constructs a wrapper for a program with various possible arguments. It is defined as part of 2 setup-hooks named `makeWrapper` and `makeBinaryWrapper` that implement the same bash functions. Hence, to use it you have to add `makeWrapper` to your `nativeBuildInputs`. Here's an example usage: + +```bash +# adds `FOOBAR=baz` to `$out/bin/foo`’s environment +makeWrapper $out/bin/foo $wrapperfile --set FOOBAR baz + +# Prefixes the binary paths of `hello` and `git` +# and suffixes the binary path of `xdg-utils`. +# Be advised that paths often should be patched in directly +# (via string replacements or in `configurePhase`). +makeWrapper $out/bin/foo $wrapperfile \ + --prefix PATH : ${lib.makeBinPath [ hello git ]} \ + --suffix PATH : ${lib.makeBinPath [ xdg-utils ]} +``` + +Packages may expect or require other utilities to be available at runtime. +`makeWrapper` can be used to add packages to a `PATH` environment variable local to a wrapper. + +Use `--prefix` to explicitly set dependencies in `PATH`. + +::: {.note} +`--prefix` essentially hard-codes dependencies into the wrapper. +They cannot be overridden without rebuilding the package. +::: + +If dependencies should be resolved at runtime, use `--suffix` to append fallback values to `PATH`. + +There’s many more kinds of arguments, they are documented in `nixpkgs/pkgs/build-support/setup-hooks/make-wrapper.sh` for the `makeWrapper` implementation and in `nixpkgs/pkgs/build-support/setup-hooks/make-binary-wrapper/make-binary-wrapper.sh` for the `makeBinaryWrapper` implementation. + +`wrapProgram` is a convenience function you probably want to use most of the time, implemented by both `makeWrapper` and `makeBinaryWrapper`. + +Using the `makeBinaryWrapper` implementation is usually preferred, as it creates a tiny _compiled_ wrapper executable, that can be used as a shebang interpreter. This is needed mostly on Darwin, where shebangs cannot point to scripts, [due to a limitation with the `execve`-syscall](https://stackoverflow.com/questions/67100831/macos-shebang-with-absolute-path-not-working). Compiled wrappers generated by `makeBinaryWrapper` can be inspected with `less <path-to-wrapper>` - by scrolling past the binary data you should be able to see the shell command that generated the executable and there see the environment variables that were injected into the wrapper. + +### `remove-references-to -t` \<storepath\> [ `-t` \<storepath\> ... ] \<file\> ... {#fun-remove-references-to} + +Removes the references of the specified files to the specified store files. This is done without changing the size of the file by replacing the hash by `eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee`, and should work on compiled executables. This is meant to be used to remove the dependency of the output on inputs that are known to be unnecessary at runtime. Of course, reckless usage will break the patched programs. +To use this, add `removeReferencesTo` to `nativeBuildInputs`. + +As `remove-references-to` is an actual executable and not a shell function, it can be used with `find`. +Example removing all references to the compiler in the output: +```nix +postInstall = '' + find "$out" -type f -exec remove-references-to -t ${stdenv.cc} '{}' + +''; +``` + +### `substitute` \<infile\> \<outfile\> \<subs\> {#fun-substitute} + +Performs string substitution on the contents of \<infile\>, writing the result to \<outfile\>. The substitutions in \<subs\> are of the following form: + +#### `--replace` \<s1\> \<s2\> {#fun-substitute-replace} + +Replace every occurrence of the string \<s1\> by \<s2\>. + +#### `--subst-var` \<varName\> {#fun-substitute-subst-var} + +Replace every occurrence of `@varName@` by the contents of the environment variable \<varName\>. This is useful for generating files from templates, using `@...@` in the template as placeholders. + +#### `--subst-var-by` \<varName\> \<s\> {#fun-substitute-subst-var-by} + +Replace every occurrence of `@varName@` by the string \<s\>. + +Example: + +```shell +substitute ./foo.in ./foo.out \ + --replace /usr/bin/bar $bar/bin/bar \ + --replace "a string containing spaces" "some other text" \ + --subst-var someVar +``` + +### `substituteInPlace` \<multiple files\> \<subs\> {#fun-substituteInPlace} + +Like `substitute`, but performs the substitutions in place on the files passed. + +### `substituteAll` \<infile\> \<outfile\> {#fun-substituteAll} + +Replaces every occurrence of `@varName@`, where \<varName\> is any environment variable, in \<infile\>, writing the result to \<outfile\>. For instance, if \<infile\> has the contents + +```bash +#! @bash@/bin/sh +PATH=@coreutils@/bin +echo @foo@ +``` + +and the environment contains `bash=/nix/store/bmwp0q28cf21...-bash-3.2-p39` and `coreutils=/nix/store/68afga4khv0w...-coreutils-6.12`, but does not contain the variable `foo`, then the output will be + +```bash +#! /nix/store/bmwp0q28cf21...-bash-3.2-p39/bin/sh +PATH=/nix/store/68afga4khv0w...-coreutils-6.12/bin +echo @foo@ +``` + +That is, no substitution is performed for undefined variables. + +Environment variables that start with an uppercase letter or an underscore are filtered out, to prevent global variables (like `HOME`) or private variables (like `__ETC_PROFILE_DONE`) from accidentally getting substituted. The variables also have to be valid bash "names", as defined in the bash manpage (alphanumeric or `_`, must not start with a number). + +### `substituteAllInPlace` \<file\> {#fun-substituteAllInPlace} + +Like `substituteAll`, but performs the substitutions in place on the file \<file\>. + +### `stripHash` \<path\> {#fun-stripHash} + +Strips the directory and hash part of a store path, outputting the name part to `stdout`. For example: + +```bash +# prints coreutils-8.24 +stripHash "/nix/store/9s9r019176g7cvn2nvcw41gsp862y6b4-coreutils-8.24" +``` + +If you wish to store the result in another variable, then the following idiom may be useful: + +```bash +name="/nix/store/9s9r019176g7cvn2nvcw41gsp862y6b4-coreutils-8.24" +someVar=$(stripHash $name) +``` + +### `wrapProgram` \<executable\> \<makeWrapperArgs\> {#fun-wrapProgram} + +Convenience function for `makeWrapper` that replaces `<\executable\>` with a wrapper that executes the original program. It takes all the same arguments as `makeWrapper`, except for `--inherit-argv0` (used by the `makeBinaryWrapper` implementation) and `--argv0` (used by both `makeWrapper` and `makeBinaryWrapper` wrapper implementations). + +If you will apply it multiple times, it will overwrite the wrapper file and you will end up with double wrapping, which should be avoided. + +### `prependToVar` \<variableName\> \<elements...\> {#fun-prependToVar} + +Prepend elements to a variable. + +Example: + +```shellSession +$ configureFlags="--disable-static" +$ prependToVar configureFlags --disable-dependency-tracking --enable-foo +$ echo $configureFlags +--disable-dependency-tracking --enable-foo --disable-static +``` + +### `appendToVar` \<variableName\> \<elements...\> {#fun-appendToVar} + +Append elements to a variable. + +Example: + +```shellSession +$ configureFlags="--disable-static" +$ appendToVar configureFlags --disable-dependency-tracking --enable-foo +$ echo $configureFlags +--disable-static --disable-dependency-tracking --enable-foo +``` + +## Package setup hooks {#ssec-setup-hooks} + +Nix itself considers a build-time dependency as merely something that should previously be built and accessible at build time—packages themselves are on their own to perform any additional setup. In most cases, that is fine, and the downstream derivation can deal with its own dependencies. But for a few common tasks, that would result in almost every package doing the same sort of setup work—depending not on the package itself, but entirely on which dependencies were used. + +In order to alleviate this burden, the setup hook mechanism was written, where any package can include a shell script that \[by convention rather than enforcement by Nix\], any downstream reverse-dependency will source as part of its build process. That allows the downstream dependency to merely specify its dependencies, and lets those dependencies effectively initialize themselves. No boilerplate mirroring the list of dependencies is needed. + +The setup hook mechanism is a bit of a sledgehammer though: a powerful feature with a broad and indiscriminate area of effect. The combination of its power and implicit use may be expedient, but isn’t without costs. Nix itself is unchanged, but the spirit of added dependencies being effect-free is violated even if the latter isn’t. For example, if a derivation path is mentioned more than once, Nix itself doesn’t care and simply makes sure the dependency derivation is already built just the same—depending is just needing something to exist, and needing is idempotent. However, a dependency specified twice will have its setup hook run twice, and that could easily change the build environment (though a well-written setup hook will therefore strive to be idempotent so this is in fact not observable). More broadly, setup hooks are anti-modular in that multiple dependencies, whether the same or different, should not interfere and yet their setup hooks may well do so. + +The most typical use of the setup hook is actually to add other hooks which are then run (i.e. after all the setup hooks) on each dependency. For example, the C compiler wrapper’s setup hook feeds itself flags for each dependency that contains relevant libraries and headers. This is done by defining a bash function, and appending its name to one of `envBuildBuildHooks`, `envBuildHostHooks`, `envBuildTargetHooks`, `envHostHostHooks`, `envHostTargetHooks`, or `envTargetTargetHooks`. These 6 bash variables correspond to the 6 sorts of dependencies by platform (there’s 12 total but we ignore the propagated/non-propagated axis). + +Packages adding a hook should not hard code a specific hook, but rather choose a variable *relative* to how they are included. Returning to the C compiler wrapper example, if the wrapper itself is an `n` dependency, then it only wants to accumulate flags from `n + 1` dependencies, as only those ones match the compiler’s target platform. The `hostOffset` variable is defined with the current dependency’s host offset `targetOffset` with its target offset, before its setup hook is sourced. Additionally, since most environment hooks don’t care about the target platform, that means the setup hook can append to the right bash array by doing something like + +```bash +addEnvHooks "$hostOffset" myBashFunction +``` + +The *existence* of setups hooks has long been documented and packages inside Nixpkgs are free to use this mechanism. Other packages, however, should not rely on these mechanisms not changing between Nixpkgs versions. Because of the existing issues with this system, there’s little benefit from mandating it be stable for any period of time. + +First, let’s cover some setup hooks that are part of Nixpkgs default `stdenv`. This means that they are run for every package built using `stdenv.mkDerivation` or when using a custom builder that has `source $stdenv/setup`. Some of these are platform specific, so they may run on Linux but not Darwin or vice-versa. + +### `move-docs.sh` {#move-docs.sh} + +This setup hook moves any installed documentation to the `/share` subdirectory directory. This includes the man, doc and info directories. This is needed for legacy programs that do not know how to use the `share` subdirectory. + +### `compress-man-pages.sh` {#compress-man-pages.sh} + +This setup hook compresses any man pages that have been installed. The compression is done using the gzip program. This helps to reduce the installed size of packages. + +### `strip.sh` {#strip.sh} + +This runs the strip command on installed binaries and libraries. This removes unnecessary information like debug symbols when they are not needed. This also helps to reduce the installed size of packages. + +### `patch-shebangs.sh` {#patch-shebangs.sh} + +This setup hook patches installed scripts to add Nix store paths to their shebang interpreter as found in the build environment. The [shebang](https://en.wikipedia.org/wiki/Shebang_(Unix)) line tells a Unix-like operating system which interpreter to use to execute the script's contents. + +::: note +The [generic builder][generic-builder] populates `PATH` from inputs of the derivation. +::: + +[generic-builder]: https://github.com/NixOS/nixpkgs/blob/19d4f7dc485f74109bd66ef74231285ff797a823/pkgs/stdenv/generic/builder.sh + +#### Invocation {#patch-shebangs.sh-invocation} + +Multiple paths can be specified. + +``` +patchShebangs [--build | --host] PATH... +``` + +##### Flags {#patch-shebangs.sh-invocation-flags} + +`--build` +: Look up commands available at build time + +`--host` +: Look up commands available at run time + +##### Examples {#patch-shebangs.sh-invocation-examples} + +```sh +patchShebangs --host /nix/store/<hash>-hello-1.0/bin +``` + +```sh +patchShebangs --build configure +``` + +`#!/bin/sh` will be rewritten to `#!/nix/store/<hash>-some-bash/bin/sh`. + +`#!/usr/bin/env` gets special treatment: `#!/usr/bin/env python` is rewritten to `/nix/store/<hash>/bin/python`. + +Interpreter paths that point to a valid Nix store location are not changed. + +::: note +A script file must be marked as executable, otherwise it will not be +considered. +::: + +This mechanism ensures that the interpreter for a given script is always found and is exactly the one specified by the build. + +It can be disabled by setting [`dontPatchShebangs`](#var-stdenv-dontPatchShebangs): + +```nix +stdenv.mkDerivation { + # ... + dontPatchShebangs = true; + # ... +} +``` + +The file [`patch-shebangs.sh`][patch-shebangs.sh] defines the [`patchShebangs`][patchShebangs] function. It is used to implement [`patchShebangsAuto`][patchShebangsAuto], the [setup hook](#ssec-setup-hooks) that is registered to run during the [fixup phase](#ssec-fixup-phase) by default. + +If you need to run `patchShebangs` at build time, it must be called explicitly within [one of the build phases](#sec-stdenv-phases). + +[patch-shebangs.sh]: https://github.com/NixOS/nixpkgs/blob/19d4f7dc485f74109bd66ef74231285ff797a823/pkgs/build-support/setup-hooks/patch-shebangs.sh +[patchShebangs]: https://github.com/NixOS/nixpkgs/blob/19d4f7dc485f74109bd66ef74231285ff797a823/pkgs/build-support/setup-hooks/patch-shebangs.sh#L24-L105 +[patchShebangsAuto]: https://github.com/NixOS/nixpkgs/blob/19d4f7dc485f74109bd66ef74231285ff797a823/pkgs/build-support/setup-hooks/patch-shebangs.sh#L107-L119 + +### `audit-tmpdir.sh` {#audit-tmpdir.sh} + +This verifies that no references are left from the install binaries to the directory used to build those binaries. This ensures that the binaries do not need things outside the Nix store. This is currently supported in Linux only. + +### `multiple-outputs.sh` {#multiple-outputs.sh} + +This setup hook adds configure flags that tell packages to install files into any one of the proper outputs listed in `outputs`. This behavior can be turned off by setting `setOutputFlags` to false in the derivation environment. See [](#chap-multiple-output) for more information. + +### `move-sbin.sh` {#move-sbin.sh} + +This setup hook moves any binaries installed in the `sbin/` subdirectory into `bin/`. In addition, a link is provided from `sbin/` to `bin/` for compatibility. + +### `move-lib64.sh` {#move-lib64.sh} + +This setup hook moves any libraries installed in the `lib64/` subdirectory into `lib/`. In addition, a link is provided from `lib64/` to `lib/` for compatibility. + +### `move-systemd-user-units.sh` {#move-systemd-user-units.sh} + +This setup hook moves any systemd user units installed in the `lib/` subdirectory into `share/`. In addition, a link is provided from `share/` to `lib/` for compatibility. This is needed for systemd to find user services when installed into the user profile. + +This hook only runs when compiling for Linux. + +### `set-source-date-epoch-to-latest.sh` {#set-source-date-epoch-to-latest.sh} + +This sets `SOURCE_DATE_EPOCH` to the modification time of the most recent file. + +### Bintools Wrapper and hook {#bintools-wrapper} + +The Bintools Wrapper wraps the binary utilities for a bunch of miscellaneous purposes. These are GNU Binutils when targeting Linux, and a mix of cctools and GNU binutils for Darwin. \[The “Bintools” name is supposed to be a compromise between “Binutils” and “cctools” not denoting any specific implementation.\] Specifically, the underlying bintools package, and a C standard library (glibc or Darwin’s libSystem, just for the dynamic loader) are all fed in, and dependency finding, hardening (see below), and purity checks for each are handled by the Bintools Wrapper. Packages typically depend on CC Wrapper, which in turn (at run time) depends on the Bintools Wrapper. + +The Bintools Wrapper was only just recently split off from CC Wrapper, so the division of labor is still being worked out. For example, it shouldn’t care about the C standard library, but just take a derivation with the dynamic loader (which happens to be the glibc on linux). Dependency finding however is a task both wrappers will continue to need to share, and probably the most important to understand. It is currently accomplished by collecting directories of host-platform dependencies (i.e. `buildInputs` and `nativeBuildInputs`) in environment variables. The Bintools Wrapper’s setup hook causes any `lib` and `lib64` subdirectories to be added to `NIX_LDFLAGS`. Since the CC Wrapper and the Bintools Wrapper use the same strategy, most of the Bintools Wrapper code is sparsely commented and refers to the CC Wrapper. But the CC Wrapper’s code, by contrast, has quite lengthy comments. The Bintools Wrapper merely cites those, rather than repeating them, to avoid falling out of sync. + +A final task of the setup hook is defining a number of standard environment variables to tell build systems which executables fulfill which purpose. They are defined to just be the base name of the tools, under the assumption that the Bintools Wrapper’s binaries will be on the path. Firstly, this helps poorly-written packages, e.g. ones that look for just `gcc` when `CC` isn’t defined yet `clang` is to be used. Secondly, this helps packages not get confused when cross-compiling, in which case multiple Bintools Wrappers may simultaneously be in use. [^footnote-stdenv-per-platform-wrapper] `BUILD_`- and `TARGET_`-prefixed versions of the normal environment variable are defined for additional Bintools Wrappers, properly disambiguating them. + +A problem with this final task is that the Bintools Wrapper is honest and defines `LD` as `ld`. Most packages, however, firstly use the C compiler for linking, secondly use `LD` anyways, defining it as the C compiler, and thirdly, only so define `LD` when it is undefined as a fallback. This triple-threat means Bintools Wrapper will break those packages, as LD is already defined as the actual linker which the package won’t override yet doesn’t want to use. The workaround is to define, just for the problematic package, `LD` as the C compiler. A good way to do this would be `preConfigure = "LD=$CC"`. + +### CC Wrapper and hook {#cc-wrapper} + +The CC Wrapper wraps a C toolchain for a bunch of miscellaneous purposes. Specifically, a C compiler (GCC or Clang), wrapped binary tools, and a C standard library (glibc or Darwin’s libSystem, just for the dynamic loader) are all fed in, and dependency finding, hardening (see below), and purity checks for each are handled by the CC Wrapper. Packages typically depend on the CC Wrapper, which in turn (at run-time) depends on the Bintools Wrapper. + +Dependency finding is undoubtedly the main task of the CC Wrapper. This works just like the Bintools Wrapper, except that any `include` subdirectory of any relevant dependency is added to `NIX_CFLAGS_COMPILE`. The setup hook itself contains elaborate comments describing the exact mechanism by which this is accomplished. + +Similarly, the CC Wrapper follows the Bintools Wrapper in defining standard environment variables with the names of the tools it wraps, for the same reasons described above. Importantly, while it includes a `cc` symlink to the c compiler for portability, the `CC` will be defined using the compiler’s “real name” (i.e. `gcc` or `clang`). This helps lousy build systems that inspect on the name of the compiler rather than run it. + +Here are some more packages that provide a setup hook. Since the list of hooks is extensible, this is not an exhaustive list. The mechanism is only to be used as a last resort, so it might cover most uses. + +### Other hooks {#stdenv-other-hooks} + +Many other packages provide hooks, that are not part of `stdenv`. You can find +these in the [Hooks Reference](#chap-hooks). + +### Compiler and Linker wrapper hooks {#compiler-linker-wrapper-hooks} + +If the file `${cc}/nix-support/cc-wrapper-hook` exists, it will be run at the end of the [compiler wrapper](#cc-wrapper). +If the file `${binutils}/nix-support/post-link-hook` exists, it will be run at the end of the linker wrapper. +These hooks allow a user to inject code into the wrappers. +As an example, these hooks can be used to extract `extraBefore`, `params` and `extraAfter` which store all the command line arguments passed to the compiler and linker respectively. + +## Purity in Nixpkgs {#sec-purity-in-nixpkgs} + +*Measures taken to prevent dependencies on packages outside the store, and what you can do to prevent them.* + +GCC doesn’t search in locations such as `/usr/include`. In fact, attempts to add such directories through the `-I` flag are filtered out. Likewise, the linker (from GNU binutils) doesn’t search in standard locations such as `/usr/lib`. Programs built on Linux are linked against a GNU C Library that likewise doesn’t search in the default system locations. + +## Hardening in Nixpkgs {#sec-hardening-in-nixpkgs} + +There are flags available to harden packages at compile or link-time. These can be toggled using the `stdenv.mkDerivation` parameters `hardeningDisable` and `hardeningEnable`. + +Both parameters take a list of flags as strings. The special `"all"` flag can be passed to `hardeningDisable` to turn off all hardening. These flags can also be used as environment variables for testing or development purposes. + +For more in-depth information on these hardening flags and hardening in general, refer to the [Debian Wiki](https://wiki.debian.org/Hardening), [Ubuntu Wiki](https://wiki.ubuntu.com/Security/Features), [Gentoo Wiki](https://wiki.gentoo.org/wiki/Project:Hardened), and the [Arch Wiki](https://wiki.archlinux.org/title/Security). + +### Hardening flags enabled by default {#sec-hardening-flags-enabled-by-default} + +The following flags are enabled by default and might require disabling with `hardeningDisable` if the program to package is incompatible. + +#### `format` {#format} + +Adds the `-Wformat -Wformat-security -Werror=format-security` compiler options. At present, this warns about calls to `printf` and `scanf` functions where the format string is not a string literal and there are no format arguments, as in `printf(foo);`. This may be a security hole if the format string came from untrusted input and contains `%n`. + +This needs to be turned off or fixed for errors similar to: + +``` +/tmp/nix-build-zynaddsubfx-2.5.2.drv-0/zynaddsubfx-2.5.2/src/UI/guimain.cpp:571:28: error: format not a string literal and no format arguments [-Werror=format-security] + printf(help_message); + ^ +cc1plus: some warnings being treated as errors +``` + +#### `stackprotector` {#stackprotector} + +Adds the `-fstack-protector-strong --param ssp-buffer-size=4` compiler options. This adds safety checks against stack overwrites rendering many potential code injection attacks into aborting situations. In the best case this turns code injection vulnerabilities into denial of service or into non-issues (depending on the application). + +This needs to be turned off or fixed for errors similar to: + +``` +bin/blib.a(bios_console.o): In function `bios_handle_cup': +/tmp/nix-build-ipxe-20141124-5cbdc41.drv-0/ipxe-5cbdc41/src/arch/i386/firmware/pcbios/bios_console.c:86: undefined reference to `__stack_chk_fail' +``` + +#### `fortify` {#fortify} + +Adds the `-O2 -D_FORTIFY_SOURCE=2` compiler options. During code generation the compiler knows a great deal of information about buffer sizes (where possible), and attempts to replace insecure unlimited length buffer function calls with length-limited ones. This is especially useful for old, crufty code. Additionally, format strings in writable memory that contain `%n` are blocked. If an application depends on such a format string, it will need to be worked around. + +Additionally, some warnings are enabled which might trigger build failures if compiler warnings are treated as errors in the package build. In this case, set `env.NIX_CFLAGS_COMPILE` to `-Wno-error=warning-type`. + +This needs to be turned off or fixed for errors similar to: + +``` +malloc.c:404:15: error: return type is an incomplete type +malloc.c:410:19: error: storage size of 'ms' isn't known + +strdup.h:22:1: error: expected identifier or '(' before '__extension__' + +strsep.c:65:23: error: register name not specified for 'delim' + +installwatch.c:3751:5: error: conflicting types for '__open_2' + +fcntl2.h:50:4: error: call to '__open_missing_mode' declared with attribute error: open with O_CREAT or O_TMPFILE in second argument needs 3 arguments +``` + +#### `pic` {#pic} + +Adds the `-fPIC` compiler options. This options adds support for position independent code in shared libraries and thus making ASLR possible. + +Most notably, the Linux kernel, kernel modules and other code not running in an operating system environment like boot loaders won’t build with PIC enabled. The compiler will is most cases complain that PIC is not supported for a specific build. + +This needs to be turned off or fixed for assembler errors similar to: + +``` +ccbLfRgg.s: Assembler messages: +ccbLfRgg.s:33: Error: missing or invalid displacement expression `private_key_len@GOTOFF' +``` + +#### `strictoverflow` {#strictoverflow} + +Signed integer overflow is undefined behaviour according to the C standard. If it happens, it is an error in the program as it should check for overflow before it can happen, not afterwards. GCC provides built-in functions to perform arithmetic with overflow checking, which are correct and faster than any custom implementation. As a workaround, the option `-fno-strict-overflow` makes gcc behave as if signed integer overflows were defined. + +This flag should not trigger any build or runtime errors. + +#### `relro` {#relro} + +Adds the `-z relro` linker option. During program load, several ELF memory sections need to be written to by the linker, but can be turned read-only before turning over control to the program. This prevents some GOT (and .dtors) overwrite attacks, but at least the part of the GOT used by the dynamic linker (.got.plt) is still vulnerable. + +This flag can break dynamic shared object loading. For instance, the module systems of Xorg and OpenCV are incompatible with this flag. In almost all cases the `bindnow` flag must also be disabled and incompatible programs typically fail with similar errors at runtime. + +#### `bindnow` {#bindnow} + +Adds the `-z bindnow` linker option. During program load, all dynamic symbols are resolved, allowing for the complete GOT to be marked read-only (due to `relro`). This prevents GOT overwrite attacks. For very large applications, this can incur some performance loss during initial load while symbols are resolved, but this shouldn’t be an issue for daemons. + +This flag can break dynamic shared object loading. For instance, the module systems of Xorg and PHP are incompatible with this flag. Programs incompatible with this flag often fail at runtime due to missing symbols, like: + +``` +intel_drv.so: undefined symbol: vgaHWFreeHWRec +``` + +### Hardening flags disabled by default {#sec-hardening-flags-disabled-by-default} + +The following flags are disabled by default and should be enabled with `hardeningEnable` for packages that take untrusted input like network services. + +#### `pie` {#pie} + +This flag is disabled by default for normal `glibc` based NixOS package builds, but enabled by default for `musl` based package builds. + +Adds the `-fPIE` compiler and `-pie` linker options. Position Independent Executables are needed to take advantage of Address Space Layout Randomization, supported by modern kernel versions. While ASLR can already be enforced for data areas in the stack and heap (brk and mmap), the code areas must be compiled as position-independent. Shared libraries already do this with the `pic` flag, so they gain ASLR automatically, but binary .text regions need to be build with `pie` to gain ASLR. When this happens, ROP attacks are much harder since there are no static locations to bounce off of during a memory corruption attack. + +Static libraries need to be compiled with `-fPIE` so that executables can link them in with the `-pie` linker option. +If the libraries lack `-fPIE`, you will get the error `recompile with -fPIE`. + +[^footnote-stdenv-ignored-build-platform]: The build platform is ignored because it is a mere implementation detail of the package satisfying the dependency: As a general programming principle, dependencies are always *specified* as interfaces, not concrete implementation. +[^footnote-stdenv-native-dependencies-in-path]: Currently, this means for native builds all dependencies are put on the `PATH`. But in the future that may not be the case for sake of matching cross: the platforms would be assumed to be unique for native and cross builds alike, so only the `depsBuild*` and `nativeBuildInputs` would be added to the `PATH`. +[^footnote-stdenv-propagated-dependencies]: Nix itself already takes a package’s transitive dependencies into account, but this propagation ensures nixpkgs-specific infrastructure like [setup hooks](#ssec-setup-hooks) also are run as if it were a propagated dependency. +[^footnote-stdenv-find-inputs-location]: The `findInputs` function, currently residing in `pkgs/stdenv/generic/setup.sh`, implements the propagation logic. +[^footnote-stdenv-sys-lib-search-path]: It clears the `sys_lib_*search_path` variables in the Libtool script to prevent Libtool from using libraries in `/usr/lib` and such. +[^footnote-stdenv-build-time-guessing-impurity]: Eventually these will be passed building natively as well, to improve determinism: build-time guessing, as is done today, is a risk of impurity. +[^footnote-stdenv-per-platform-wrapper]: Each wrapper targets a single platform, so if binaries for multiple platforms are needed, the underlying binaries must be wrapped multiple times. As this is a property of the wrapper itself, the multiple wrappings are needed whether or not the same underlying binaries can target multiple platforms. |