You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
nixpkgs/pkgs/stdenv/linux/make-bootstrap-tools-cross.nix

283 lines
9.1 KiB

{system ? builtins.currentSystem}:
let buildFor = toolsArch: (
let
pkgsFun = import ../../..;
pkgsNoParams = pkgsFun {};
sheevaplugCrossSystem = {
crossSystem = rec {
config = "arm-linux-gnueabi";
bigEndian = false;
arch = "armv5te";
float = "soft";
withTLS = true;
libc = "glibc";
platform = pkgsNoParams.platforms.sheevaplug;
openssl.system = "linux-generic32";
inherit (platform) gcc;
};
};
raspberrypiCrossSystem = {
crossSystem = rec {
config = "arm-linux-gnueabihf";
bigEndian = false;
arch = "armv6";
float = "hard";
fpu = "vfp";
withTLS = true;
libc = "glibc";
platform = pkgsNoParams.platforms.raspberrypi;
openssl.system = "linux-generic32";
inherit (platform) gcc;
};
};
armv7l-hf-multiplatform-crossSystem = {
crossSystem = rec {
config = "arm-linux-gnueabihf";
bigEndian = false;
arch = "armv7-a";
float = "hard";
fpu = "vfpv3-d16";
withTLS = true;
libc = "glibc";
platform = pkgsNoParams.platforms.armv7l-hf-multiplatform;
openssl.system = "linux-generic32";
inherit (platform) gcc;
};
};
aarch64-multiplatform-crossSystem = {
crossSystem = rec {
config = "aarch64-linux-gnu";
bigEndian = false;
arch = "aarch64";
withTLS = true;
libc = "glibc";
platform = pkgsNoParams.platforms.aarch64-multiplatform;
inherit (platform) gcc;
};
};
selectedCrossSystem =
if toolsArch == "armv5tel" then sheevaplugCrossSystem else
if toolsArch == "armv6l" then raspberrypiCrossSystem else
if toolsArch == "armv7l" then armv7l-hf-multiplatform-crossSystem else
if toolsArch == "aarch64" then aarch64-multiplatform-crossSystem else 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
6 years ago
pkgsUnspliced = pkgsFun ({inherit system;} // selectedCrossSystem);
pkgs = pkgsUnspliced.splicedPackages;
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
6 years ago
inherit (pkgsUnspliced.buildPackages) stdenv nukeReferences cpio binutilsCross;
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
6 years ago
glibc = pkgs.libcCross.nativeDrv;
bash = pkgs.bash.crossDrv;
findutils = pkgs.findutils.crossDrv;
diffutils = pkgs.diffutils.crossDrv;
gnused = pkgs.gnused.crossDrv;
gnugrep = pkgs.gnugrep.crossDrv;
gawk = pkgs.gawk.crossDrv;
gzip = pkgs.gzip.crossDrv;
bzip2 = pkgs.bzip2.crossDrv;
gnumake = pkgs.gnumake.crossDrv;
patch = pkgs.patch.crossDrv;
patchelf = pkgs.patchelf.crossDrv;
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
6 years ago
gcc = pkgs.gcc.crossDrv.cc;
gmpxx = pkgs.gmpxx.crossDrv;
mpfr = pkgs.mpfr.crossDrv;
zlib = pkgs.zlib.crossDrv;
libmpc = pkgs.libmpc.crossDrv;
binutils = pkgs.binutils.crossDrv;
libelf = pkgs.libelf.crossDrv;
# Keep these versions in sync with the versions used in the current GCC!
isl = pkgs.isl_0_14.crossDrv;
in
rec {
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
6 years ago
coreutilsMinimal = pkgs.coreutils.override (args: {
# We want coreutils without ACL/attr support.
aclSupport = false;
attrSupport = false;
# Our tooling currently can't handle scripts in bin/, only ELFs and symlinks.
singleBinary = "symlinks";
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
6 years ago
});
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
6 years ago
tarMinimal = pkgs.gnutar.override { acl = 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
6 years ago
busyboxMinimal = pkgs.busybox.override {
useMusl = true;
enableStatic = true;
enableMinimal = true;
extraConfig = ''
CONFIG_ASH y
CONFIG_ASH_BUILTIN_ECHO y
CONFIG_ASH_BUILTIN_TEST y
CONFIG_ASH_OPTIMIZE_FOR_SIZE y
CONFIG_MKDIR y
CONFIG_TAR y
CONFIG_UNXZ y
'';
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
6 years ago
};
build =
stdenv.mkDerivation {
name = "stdenv-bootstrap-tools-cross";
crossConfig = pkgsUnspliced.hostPlatform.config;
buildInputs = [nukeReferences cpio binutilsCross];
buildCommand = ''
set -x
mkdir -p $out/bin $out/lib $out/libexec
# Copy what we need of Glibc.
cp -d ${glibc.out}/lib/ld-*.so* $out/lib
cp -d ${glibc.out}/lib/libc*.so* $out/lib
cp -d ${glibc.out}/lib/libc_nonshared.a $out/lib
cp -d ${glibc.out}/lib/libm*.so* $out/lib
cp -d ${glibc.out}/lib/libdl*.so* $out/lib
cp -d ${glibc.out}/lib/librt*.so* $out/lib
cp -d ${glibc.out}/lib/libpthread*.so* $out/lib
cp -d ${glibc.out}/lib/libnsl*.so* $out/lib
cp -d ${glibc.out}/lib/libutil*.so* $out/lib
cp -d ${glibc.out}/lib/libnss*.so* $out/lib
cp -d ${glibc.out}/lib/libresolv*.so* $out/lib
cp -d ${glibc.out}/lib/crt?.o $out/lib
cp -rL ${glibc.dev}/include $out
chmod -R u+w "$out"
# glibc can contain linker scripts: find them, copy their deps,
# and get rid of absolute paths (nuke-refs would make them useless)
local lScripts=$(grep --files-with-matches --max-count=1 'GNU ld script' -R "$out/lib")
cp -d -t "$out/lib/" $(cat $lScripts | tr " " "\n" | grep -F '${glibc.out}' | sort -u)
for f in $lScripts; do
substituteInPlace "$f" --replace '${glibc.out}/lib/' ""
done
# Hopefully we won't need these.
rm -rf $out/include/mtd $out/include/rdma $out/include/sound $out/include/video
find $out/include -name .install -exec rm {} \;
find $out/include -name ..install.cmd -exec rm {} \;
mv $out/include $out/include-glibc
# Copy coreutils, bash, etc.
cp ${coreutilsMinimal}/bin/* $out/bin
(cd $out/bin && rm vdir dir sha*sum pinky factor pathchk runcon shuf who whoami shred users)
cp ${bash}/bin/bash $out/bin
cp ${findutils}/bin/find $out/bin
cp ${findutils}/bin/xargs $out/bin
cp -d ${diffutils}/bin/* $out/bin
cp -d ${gnused}/bin/* $out/bin
cp -d ${gnugrep}/bin/grep $out/bin
cp ${gawk}/bin/gawk $out/bin
cp -d ${gawk}/bin/awk $out/bin
cp ${tarMinimal}/bin/tar $out/bin
cp ${gzip}/bin/gzip $out/bin
cp ${bzip2.bin}/bin/bzip2 $out/bin
cp -d ${gnumake}/bin/* $out/bin
cp -d ${patch}/bin/* $out/bin
cp ${patchelf}/bin/* $out/bin
cp -d ${gnugrep.pcre.crossDrv.out}/lib/libpcre*.so* $out/lib # needed by grep
# Copy what we need of GCC.
cp -d ${gcc.out}/bin/gcc $out/bin
cp -d ${gcc.out}/bin/cpp $out/bin
cp -d ${gcc.out}/bin/g++ $out/bin
cp -d ${gcc.lib}/lib*/libgcc_s.so* $out/lib
cp -d ${gcc.lib}/lib*/libstdc++.so* $out/lib
cp -rd ${gcc.out}/lib/gcc $out/lib
chmod -R u+w $out/lib
rm -f $out/lib/gcc/*/*/include*/linux
rm -f $out/lib/gcc/*/*/include*/sound
rm -rf $out/lib/gcc/*/*/include*/root
rm -f $out/lib/gcc/*/*/include-fixed/asm
rm -rf $out/lib/gcc/*/*/plugin
#rm -f $out/lib/gcc/*/*/*.a
cp -rd ${gcc.out}/libexec/* $out/libexec
chmod -R u+w $out/libexec
rm -rf $out/libexec/gcc/*/*/plugin
mkdir $out/include
cp -rd ${gcc.out}/include/c++ $out/include
chmod -R u+w $out/include
rm -rf $out/include/c++/*/ext/pb_ds
rm -rf $out/include/c++/*/ext/parallel
cp -d ${gmpxx.out}/lib/libgmp*.so* $out/lib
cp -d ${mpfr.out}/lib/libmpfr*.so* $out/lib
cp -d ${libmpc.out}/lib/libmpc*.so* $out/lib
cp -d ${zlib.out}/lib/libz.so* $out/lib
cp -d ${libelf}/lib/libelf.so* $out/lib
# These needed for cross but not native tools because the stdenv
# GCC has certain things built in statically. See
# pkgs/stdenv/linux/default.nix for the details.
cp -d ${isl}/lib/libisl*.so* $out/lib
# Also this is needed since bzip2 uses a custom build system
# for native builds but autoconf (via a patch) for cross builds
cp -d ${bzip2.out}/lib/libbz2.so* $out/lib
# Copy binutils.
for i in as ld ar ranlib nm strip readelf objdump; do
cp ${binutils.out}/bin/$i $out/bin
done
cp -d ${binutils.out}/lib/lib*.so* $out/lib
chmod -R u+w $out
# Strip executables even further.
for i in $out/bin/* $out/libexec/gcc/*/*/*; do
if test -x $i -a ! -L $i; then
chmod +w $i
$crossConfig-strip -s $i || true
fi
done
nuke-refs $out/bin/*
nuke-refs $out/lib/*
nuke-refs $out/libexec/gcc/*/*/*
mkdir $out/.pack
mv $out/* $out/.pack
mv $out/.pack $out/pack
mkdir $out/on-server
tar cvfJ $out/on-server/bootstrap-tools.tar.xz --hard-dereference --sort=name --numeric-owner --owner=0 --group=0 --mtime=@1 -C $out/pack .
cp ${busyboxMinimal}/bin/busybox $out/on-server
chmod u+w $out/on-server/busybox
nuke-refs $out/on-server/busybox
''; # */
# The result should not contain any references (store paths) so
# that we can safely copy them out of the store and to other
# locations in the store.
allowedReferences = [];
};
dist = stdenv.mkDerivation {
name = "stdenv-bootstrap-tools-cross";
buildCommand = ''
mkdir -p $out/nix-support
echo "file tarball ${build}/on-server/bootstrap-tools.tar.xz" >> $out/nix-support/hydra-build-products
echo "file busybox ${build}/on-server/busybox" >> $out/nix-support/hydra-build-products
'';
};
}
); in {
armv5tel = buildFor "armv5tel";
armv6l = buildFor "armv6l";
armv7l = buildFor "armv7l";
aarch64 = buildFor "aarch64";
}