nixpkgs-suyu/pkgs/top-level/stage.nix

135 lines
4.7 KiB
Nix
Raw Normal View History

/* This file composes a single bootstrapping stage of the Nix Packages
collection. That is, it imports the functions that build the various
packages, and calls them with appropriate arguments. The result is a set of
all the packages in the Nix Packages collection for some particular platform
for some particular stage.
Default arguments are only provided for bootstrapping
arguments. Normal users should not import this directly but instead
import `pkgs/default.nix` or `default.nix`. */
{ ## Misc parameters kept the same for all stages
##
# Utility functions, could just import but passing in for efficiency
lib
, # Use to reevaluate Nixpkgs; a dirty hack that should be removed
nixpkgsFun
## Other parameters
##
, # The package set used at build-time. If null, `buildPackages` will
# be defined internally as the final produced package set itself. This allows
# us to avoid expensive splicing.
top-level: Introduce `buildPackages` for resolving build-time deps [N.B., this package also applies to the commits that follow it in the same PR.] In most cases, buildPackages = pkgs so things work just as before. For cross compiling, however, buildPackages is resolved as the previous bootstrapping stage. This allows us to avoid the mkDerivation hacks cross compiling currently uses today. To avoid a massive refactor, callPackage will splice together both package sets. Again to avoid churn, it uses the old `nativeDrv` vs `crossDrv` to do so. So now, whether cross compiling or not, packages with get a `nativeDrv` and `crossDrv`---in the non-cross-compiling case they are simply the same derivation. This is good because it reduces the divergence between the cross and non-cross dataflow. See `pkgs/top-level/splice.nix` for a comment along the lines of the preceding paragraph, and the code that does this splicing. Also, `forceNativeDrv` is replaced with `forceNativePackages`. The latter resolves `pkgs` unless the host platform is different from the build platform, in which case it resolves to `buildPackages`. Note that the target platform is not important here---it will not prevent `forcedNativePackages` from resolving to `pkgs`. -------- Temporarily, we make preserve some dubious decisions in the name of preserving hashes: Most importantly, we don't distinguish between "host" and "target" in the autoconf sense. This leads to the proliferation of *Cross derivations currently used. What we ought to is resolve native deps of the cross "build packages" (build = host != target) package set against the "vanilla packages" (build = host = target) package set. Instead, "build packages" uses itself, with (informally) target != build in all cases. This is wrong because it violates the "sliding window" principle of bootstrapping stages that shifting the platform triple of one stage to the left coincides with the next stage's platform triple. Only because we don't explicitly distinguish between "host" and "target" does it appear that the "sliding window" principle is preserved--indeed it is over the reductionary "platform double" of just "build" and "host/target". Additionally, we build libc, libgcc, etc in the same stage as the compilers themselves, which is wrong because they are used at runtime, not build time. Fixing this is somewhat subtle, and the solution and problem will be better explained in the commit that does fix it. Commits after this will solve both these issues, at the expense of breaking cross hashes. Native hashes won't be broken, thankfully. -------- Did the temporary ugliness pan out? Of the packages that currently build in `release-cross.nix`, the only ones that have their hash changed are `*.gcc.crossDrv` and `bootstrapTools.*.coreutilsMinimal`. In both cases I think it doesn't matter. 1. GCC when doing a `build = host = target = foreign` build (maximally cross), still defines environment variables like `CPATH`[1] with packages. This seems assuredly wrong because whether gcc dynamically links those, or the programs built by gcc dynamically link those---I have no idea which case is reality---they should be foreign. Therefore, in all likelihood, I just made the gcc less broken. 2. Coreutils (ab)used the old cross-compiling infrastructure to depend on a native version of itself. When coreutils was overwritten to be built with fewer features, the native version it used would also be overwritten because the binding was tight. Now it uses the much looser `BuildPackages.coreutils` which is just fine as a richer build dep doesn't cause any problems and avoids a rebuild. So, in conclusion I'd say the conservatism payed off. Onward to actually raking the muck in the next PR! [1]: https://gcc.gnu.org/onlinedocs/gcc/Environment-Variables.html
2016-12-18 08:51:18 +01:00
buildPackages
, # The package set used in the next stage. If null, `targetPackages` will be
# defined internally as the final produced package set itself, just like with
# `buildPackages` and for the same reasons.
#
# THIS IS A HACK for compilers that don't think critically about cross-
# compilation. Please do *not* use unless you really know what you are doing.
targetPackages
, # The standard environment to use for building packages.
stdenv
, # This is used because stdenv replacement and the stdenvCross do benefit from
# the overridden configuration provided by the user, as opposed to the normal
# bootstrapping stdenvs.
allowCustomOverrides
, # Non-GNU/Linux OSes are currently "impure" platforms, with their libc
# outside of the store. Thus, GCC, GFortran, & co. must always look for files
# in standard system directories (/usr/include, etc.)
noSysDirs ? stdenv.buildPlatform.system != "x86_64-freebsd"
&& stdenv.buildPlatform.system != "i686-freebsd"
&& stdenv.buildPlatform.system != "x86_64-solaris"
&& stdenv.buildPlatform.system != "x86_64-kfreebsd-gnu"
, # The configuration attribute set
config
, # A list of overlays (Additional `self: super: { .. }` customization
# functions) to be fixed together in the produced package set
overlays
}:
let
stdenvAdapters = self: super:
let res = import ../stdenv/adapters.nix self; in res // {
stdenvAdapters = res;
};
trivialBuilders = self: super:
import ../build-support/trivial-builders.nix {
inherit lib; inherit (self) stdenv stdenvNoCC; inherit (self.xorg) lndir;
};
stdenvBootstappingAndPlatforms = self: super: {
buildPackages = (if buildPackages == null then self else buildPackages)
// { recurseForDerivations = false; };
targetPackages = (if targetPackages == null then self else targetPackages)
// { recurseForDerivations = false; };
inherit stdenv;
};
# The old identifiers for cross-compiling. These should eventually be removed,
# and the packages that rely on them refactored accordingly.
platformCompat = self: super: let
inherit (super.stdenv) buildPlatform hostPlatform targetPlatform;
in {
stdenv = super.stdenv // {
inherit (super.stdenv.buildPlatform) platform;
};
inherit buildPlatform hostPlatform targetPlatform;
inherit (buildPlatform) system platform;
};
splice = self: super: import ./splice.nix lib self (buildPackages != null);
top-level: Introduce `buildPackages` for resolving build-time deps [N.B., this package also applies to the commits that follow it in the same PR.] In most cases, buildPackages = pkgs so things work just as before. For cross compiling, however, buildPackages is resolved as the previous bootstrapping stage. This allows us to avoid the mkDerivation hacks cross compiling currently uses today. To avoid a massive refactor, callPackage will splice together both package sets. Again to avoid churn, it uses the old `nativeDrv` vs `crossDrv` to do so. So now, whether cross compiling or not, packages with get a `nativeDrv` and `crossDrv`---in the non-cross-compiling case they are simply the same derivation. This is good because it reduces the divergence between the cross and non-cross dataflow. See `pkgs/top-level/splice.nix` for a comment along the lines of the preceding paragraph, and the code that does this splicing. Also, `forceNativeDrv` is replaced with `forceNativePackages`. The latter resolves `pkgs` unless the host platform is different from the build platform, in which case it resolves to `buildPackages`. Note that the target platform is not important here---it will not prevent `forcedNativePackages` from resolving to `pkgs`. -------- Temporarily, we make preserve some dubious decisions in the name of preserving hashes: Most importantly, we don't distinguish between "host" and "target" in the autoconf sense. This leads to the proliferation of *Cross derivations currently used. What we ought to is resolve native deps of the cross "build packages" (build = host != target) package set against the "vanilla packages" (build = host = target) package set. Instead, "build packages" uses itself, with (informally) target != build in all cases. This is wrong because it violates the "sliding window" principle of bootstrapping stages that shifting the platform triple of one stage to the left coincides with the next stage's platform triple. Only because we don't explicitly distinguish between "host" and "target" does it appear that the "sliding window" principle is preserved--indeed it is over the reductionary "platform double" of just "build" and "host/target". Additionally, we build libc, libgcc, etc in the same stage as the compilers themselves, which is wrong because they are used at runtime, not build time. Fixing this is somewhat subtle, and the solution and problem will be better explained in the commit that does fix it. Commits after this will solve both these issues, at the expense of breaking cross hashes. Native hashes won't be broken, thankfully. -------- Did the temporary ugliness pan out? Of the packages that currently build in `release-cross.nix`, the only ones that have their hash changed are `*.gcc.crossDrv` and `bootstrapTools.*.coreutilsMinimal`. In both cases I think it doesn't matter. 1. GCC when doing a `build = host = target = foreign` build (maximally cross), still defines environment variables like `CPATH`[1] with packages. This seems assuredly wrong because whether gcc dynamically links those, or the programs built by gcc dynamically link those---I have no idea which case is reality---they should be foreign. Therefore, in all likelihood, I just made the gcc less broken. 2. Coreutils (ab)used the old cross-compiling infrastructure to depend on a native version of itself. When coreutils was overwritten to be built with fewer features, the native version it used would also be overwritten because the binding was tight. Now it uses the much looser `BuildPackages.coreutils` which is just fine as a richer build dep doesn't cause any problems and avoids a rebuild. So, in conclusion I'd say the conservatism payed off. Onward to actually raking the muck in the next PR! [1]: https://gcc.gnu.org/onlinedocs/gcc/Environment-Variables.html
2016-12-18 08:51:18 +01:00
allPackages = self: super:
let res = import ./all-packages.nix
{ inherit lib nixpkgsFun noSysDirs config; }
res self;
in res;
aliases = self: super: import ./aliases.nix super;
# stdenvOverrides is used to avoid having multiple of versions
# of certain dependencies that were used in bootstrapping the
# standard environment.
stdenvOverrides = self: super:
(super.stdenv.overrides or (_: _: {})) self super;
# Allow packages to be overridden globally via the `packageOverrides'
# configuration option, which must be a function that takes `pkgs'
# as an argument and returns a set of new or overridden packages.
# The `packageOverrides' function is called with the *original*
# (un-overridden) set of packages, allowing packageOverrides
# attributes to refer to the original attributes (e.g. "foo =
# ... pkgs.foo ...").
configOverrides = self: super:
lib.optionalAttrs allowCustomOverrides
((config.packageOverrides or (super: {})) super);
# The complete chain of package set builders, applied from top to bottom
toFix = lib.foldl' (lib.flip lib.extends) (self: {}) ([
stdenvBootstappingAndPlatforms
platformCompat
stdenvAdapters
trivialBuilders
top-level: Introduce `buildPackages` for resolving build-time deps [N.B., this package also applies to the commits that follow it in the same PR.] In most cases, buildPackages = pkgs so things work just as before. For cross compiling, however, buildPackages is resolved as the previous bootstrapping stage. This allows us to avoid the mkDerivation hacks cross compiling currently uses today. To avoid a massive refactor, callPackage will splice together both package sets. Again to avoid churn, it uses the old `nativeDrv` vs `crossDrv` to do so. So now, whether cross compiling or not, packages with get a `nativeDrv` and `crossDrv`---in the non-cross-compiling case they are simply the same derivation. This is good because it reduces the divergence between the cross and non-cross dataflow. See `pkgs/top-level/splice.nix` for a comment along the lines of the preceding paragraph, and the code that does this splicing. Also, `forceNativeDrv` is replaced with `forceNativePackages`. The latter resolves `pkgs` unless the host platform is different from the build platform, in which case it resolves to `buildPackages`. Note that the target platform is not important here---it will not prevent `forcedNativePackages` from resolving to `pkgs`. -------- Temporarily, we make preserve some dubious decisions in the name of preserving hashes: Most importantly, we don't distinguish between "host" and "target" in the autoconf sense. This leads to the proliferation of *Cross derivations currently used. What we ought to is resolve native deps of the cross "build packages" (build = host != target) package set against the "vanilla packages" (build = host = target) package set. Instead, "build packages" uses itself, with (informally) target != build in all cases. This is wrong because it violates the "sliding window" principle of bootstrapping stages that shifting the platform triple of one stage to the left coincides with the next stage's platform triple. Only because we don't explicitly distinguish between "host" and "target" does it appear that the "sliding window" principle is preserved--indeed it is over the reductionary "platform double" of just "build" and "host/target". Additionally, we build libc, libgcc, etc in the same stage as the compilers themselves, which is wrong because they are used at runtime, not build time. Fixing this is somewhat subtle, and the solution and problem will be better explained in the commit that does fix it. Commits after this will solve both these issues, at the expense of breaking cross hashes. Native hashes won't be broken, thankfully. -------- Did the temporary ugliness pan out? Of the packages that currently build in `release-cross.nix`, the only ones that have their hash changed are `*.gcc.crossDrv` and `bootstrapTools.*.coreutilsMinimal`. In both cases I think it doesn't matter. 1. GCC when doing a `build = host = target = foreign` build (maximally cross), still defines environment variables like `CPATH`[1] with packages. This seems assuredly wrong because whether gcc dynamically links those, or the programs built by gcc dynamically link those---I have no idea which case is reality---they should be foreign. Therefore, in all likelihood, I just made the gcc less broken. 2. Coreutils (ab)used the old cross-compiling infrastructure to depend on a native version of itself. When coreutils was overwritten to be built with fewer features, the native version it used would also be overwritten because the binding was tight. Now it uses the much looser `BuildPackages.coreutils` which is just fine as a richer build dep doesn't cause any problems and avoids a rebuild. So, in conclusion I'd say the conservatism payed off. Onward to actually raking the muck in the next PR! [1]: https://gcc.gnu.org/onlinedocs/gcc/Environment-Variables.html
2016-12-18 08:51:18 +01:00
splice
allPackages
aliases
stdenvOverrides
configOverrides
] ++ overlays);
in
# Return the complete set of packages.
lib.fix toFix