Now gcc is just another build input, making it possible in the future
to have a stdenv that doesn't depend on a C compiler. This is very
useful on NixOS, since it would allow trivial builders like
writeTextFile to work without pulling in the C compiler.
If $src refers to a directory, then always copy it. Previously, we
checked the extension first, so if the directory had an extension like
.tar, unpackPhase would fail.
If a build input is a regular file, use it as a setup hook. This makes
setup hooks more efficient to create: you don't need a derivation that
copies them to $out/nix-support/setup-hook, instead you can use the
file as is.
You can now register multiple values per named hook, e.g.
addHook preConfigure "echo foo"
addHook preConfigure "echo bar"
will cause ‘runHook preConfigure’ to run both ‘echo foo’ and ‘echo
bar’ (in that order). It will also call the shell function
preConfigure() or eval the shell variable $preConfigure, if
defined. Thus, if you don't call addHook, it works like the old hook
mechanism.
Allowing multiple hooks makes stdenv more modular and extensible. For
instance, multiple setup hooks can define a preFixup hook, and all of
these will be executed.
This variable sets the minimal Mac OS X version required for
running binaries produced by the Darwin toolchain. Since it
defaults to the version of the user's SDK, setting it explicitly
should make our builds more deterministic. It's now set to 10.6
because that's what hydra.nixos.org runs.
Commit 262c21ed46 purported to enable
ignoreNulls, but it was bogus because it set the flag on the wrong
derivation (i.e. stdenv rather than the result of mkDerivation).
Recent versions of Xcode don't install headers in /usr/include but
in a directory like
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include
So use that instead, falling back to /usr/include in case of an older
version of Xcode.
Commit 986f361946 started to use
<nix/fetchurl.nix> to "download" the bootstrap binaries from the
Nixpkgs tree, using the file:/// scheme. This has really bad
consequences:
* It makes any derivation depend on the path of the Nixpkgs tree. So
evaluating a package will produce a different .drv file when run
from different locations. No wonder Hydra evaluation has been so
slow lately: for every Nixpkgs evaluation, it had to create tens of
thousands of .drv files, even if nothing had changed.
* It requires the builder to have file system access to the Nixpkgs
tree. So if your tree is in your home directory, the stdenv
bootstrap would probably fail.
So now the binaries are downloaded from tarballs.nixos.org.
Also dropped PowerPC "support".
Stdenv adapters are kinda weird and un-idiomatic (especially when they
don't actually change stdenv). It's more idiomatic to say
buildInputs = [ makeCoverageAnalysisReport ];
This removes the need for hacks like stdenv.regenerate. It also
ensures that overrideGCC is now stackable (so ‘stdenv = useGoldLinker
clangStdenv’ works).
setup.sh uses the anti-pattern `for f in $(find ...); do` in several
places. `find` returns one path per line, but `for` splits its arguments
by words, so paths which contain spaces are incorrectly split! The
correct way is `find ... | while read f; do`
Binutils nowadays contains ld.gold, which depends on libstdc++. So it
needs to be built with the new GCC rather than the one from
bootstrap-tools.
Issue #1469.
Treating fixupPhase specially is really ugly. Also, it collides with
the work in the multiple-outputs branch (which already has support for
fixing up all outputs).
Partial revert of 0a44a09121.
Some programs, e.g. guile-config, has a shebang that ends in '\':
#!/usr/bin/guile-1.8 \
-e main -s
!#
;;;; guile-config --- utility for linking programs with Guile
;;;; Jim Blandy <jim@red-bean.com> --- September 1997
This currently breaks patchShebangs:
$ read oldPath arg0 args <<< 'shebang \'; echo $?
1
$ echo $oldPath
shebang
$ echo $arg0
$ echo $args
(And setup.sh/patchShebangs is run with 'set -e' so any command that
return non-zero aborts the build.)
Fix by telling 'read' to not interpret backslashes (with the -r flag):
$ read -r oldPath arg0 args <<< 'shebang \'; echo $?
0
$ echo $oldPath
shebang
$ echo $arg0
\
$ echo $args
Also needed: escape the escape characters so that sed doesn't interpret
them.
patchShebangs has a bug that shows itself on files that have the
executable bit set but have no shebang (i.e. a blank/empty first line).
The shell would then evaluate this:
if [ != '#!' ]; then
# not evaluated
fi
With proper quoting we get the correct behaviour:
if [ "" != '#!' ]; then
# this will be evaluated
fi
When building e.g. perl for the first time there is no perl in PATH yet,
so command -v perl will fail.
This brings back the previous behaviour of silently not patching
shebangs for which there is no available command in PATH.
Currently "/usr/bin/env python" is rewritten to "/nix/store/.../env
python". That doesn't really improve anything because the interpreter
still have to be located in $PATH at runtime. The result is that many
nix package expressions do .../bin/env fixup themselves.
Instead of everyone having to do this patching locally, add the
functionality to the standard environment patchShebangs function so that
everyone can benefit.
The function ‘mkDerivation’ now checks whether the current platform
type is included in a package's meta.platform field. If not, it
throws an exception:
$ nix-build -A linux --argstr system x86_64-darwin
error: user-thrown exception: the package ‘linux-3.10.15’ is not supported on ‘x86_64-darwin’
These packages also no longer show up in ‘nix-env -qa’ output. This
means, for instance, that the number of packages shown on
x86_64-freebsd has dropped from 9268 to 4764.
Since meta.platforms was also used to prevent Hydra from building some
packages, there now is a new attribute meta.hydraPlatforms listing the
platforms on which Hydra should build the package (which defaults to
meta.platforms).
meta.license is can be a string or a list of strings. But there is one
unhandled case where "unfree" (or "unfree-redistributable") is a part of
a list. It will currently not be detected as an "unfree" package and
Hydra will attempt to build it. This should fix it.
Example: http://hydra.nixos.org/build/6553461
GCC provides a number of libraries that are used by programs built by
GCC, in particular libgcc_s.so and libstdc++.so. This caused programs
that used these libraries to have a runtime dependency on all of GCC
(~77 MiB). Now they only depend on the "lib" output of GCC (~1.6
MiB).
With this and previous multiple-output improvements, closure sizes are
reduced a lot:
hello: 41 MiB -> 22 MiB
patchelf: 118 MiB -> 23 MiB
pan: 364 MiB -> 90 MiB
This reduces the size of closures a lot by getting rid of Glibc's
header files, static libraries and utilities. For instance, the
closure of GNU Hello went from 41 MiB to 22 MiB.
TODO: We may want to merge back in parts of Glibc that we previously
split off (such as the info files).
set CMAKE_LIBRARY_PATH, CMAKE_INCLUDE_PATH based on NIX_CFLAGS_COMPILE and
NIX_LDFLAGS so that cmake's find_library like functions find all the libraries
gcc knows about thanks to the gcc wrapper
This is particular useful with myEnvFun which then also sets those CMAKE_* env
variables.`
Because setup.sh has to change this causes many rebuilds - thus it should be
included in a stdenv-update like branch
Also cmake builds in parallel perfectly fine
update cmake to latest minor number, I didn't change the patches,
just reapplied them manually recordin a new patch
Since nix-1.4, nix's corepkgs contain a fetchurl suitable for
downloading the bootstrap binaries. Doing this will allow us to have a
nixpkgs with no in-tree binaries without breaking the purity of the
bootstrap (though for now, they are fetched in-tree until the binaries
are added to nixos.org somewhere). As an additional small benefit, the
in-tree binaries do not have to be hashed on every instantiation as they
do now.
The fetchurl in nix-1.2 and 1.3 isn't able to make binaries executable,
so it can't be used for this case. In that case, attempting to build the
bootstrap will show a message asking the user to manually download each
file and add it to the store with nix-store --add (but the hash is
ultimately the same, of course).
Signed-off-by: Shea Levy <shea@shealevy.com>
Conflict in kerberos, which was updated both in master and in
stdenv-updates. Kept the stdenv-updates version, except pulled in the
enableParallelBuilding change from master.
Signed-off-by: Shea Levy <shea@shealevy.com>
Conflicts:
pkgs/development/libraries/kerberos/krb5.nix
Before this, the passthru attributes were only merged in with the
derivation attribute set, and there was no way to distinguish after the
fact which attributes were part of the derivation and which came from
passthru. Now passthru can be looked at separately as well.
Signed-off-by: Shea Levy <shea@shealevy.com>
Conflicts:
pkgs/development/libraries/libxslt/default.nix
Commit 1764ea2b0a introduced changes to libxslt
in an awkward way to avoid re-builds on Linux. This patch has been simplified
during this merge.
With multiple outputs, adding attributes to a derivation without
changing the {drv,out}Path is no longer as trivial as simply using the
`//' operator, as we usually want to add the attribute to _each_ output,
and even if we only care about one that one output can be reached via
multiple paths.
For stdenv.mkDerivation, we already had code in place to add passthru
and meta attributes to derivations. This commit simply factors part of
that code out into a lib function addPassthru, which takes a derivation
and an attribute set and appends the attribute set to each output of the
derivation.
Signed-off-by: Shea Levy <shea@shealevy.com>
Before, only the first output (and not even that when accessed through 'all' or its corresponding attribtue) had meta information and the relevant passthru attributes.
This doesn't change stdenv's hash and the tarball still builds, I'm pretty sure this is safe for master.
I'm not sure whether this was by intention, but so far postPatch hooks were
silently skipped whenever the patches list was empty. This change could possibly
change the build results of the following packages:
* gcc
* cmake (264)
* systemtap
* quemu-kvm
These packages all have in common that they have a postPatch hook and the
patches list can be empty when certain conditions are met.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Due to xz being override in the last stdenv and also in the previous, the
nixpkgs xz ended up being built by bootstrap-tools, and thus depending on it
through libgcc_so.so.1. That ends up making 'nix' with a runtime
dependency on bootstrap-tools.
If set to false, mkDerivation will throw an exception if a package has
an unfree license. ‘release-lib.nix’ uses this to enforce that we
don't build unfree packages as part of the Nixpkgs channel. Since
this is set through Nixpkgs' ‘config’ argument, it's more finegrained
than $HYDRA_DISALLOW_UNFREE.
Although patching it made some programs run (configure tests), some others
crashed with segfault. So I don't think there is any win patching it. The
proper way to solve the bootstrap in the raspberry pi is, as far as I've been
testing, use glibc 2.17 libs in bootstrap-tools with the same ld.so name as the
bootstrapped glibc.
This is a problem inherent in our way to bootstrap, that first replaces
the glibc of a given gcc+glibc (bootstrap-tools) with gcc-wrapper tricks, and
then builds a new gcc. A nicer way would be to build a gcc without glibc,
then the glibc, then the final gcc, as we do with cross-tools.
Some comments about this problem in
https://github.com/NixOS/nixpkgs/issues/234#issuecomment-11764352
This solves the problem of the change of name in ld.so between
glibc 2.13 and 2.17 (at least for armhf). Some comments about it in
https://github.com/NixOS/nixpkgs/issues/234#issuecomment-11764352
As there says, without this, gmp (after glibc built) tests crashed
- segfault.
Now that we have a way to alter /bin/sh in chroots on non-NixOS Linux
platforms, it may be useful to have access to stdenv.shell package and to
static bash contained in bootstrap tools. So make them accessible via stdenv
attribute set.
libstdc++ and libmudflapth link to the dynamic loader; if the
bootstrap uses another dynamic loader name, and the rpath of these
libs isn't changed, they will fail to load.
Conflicts:
pkgs/development/compilers/gcc/4.6/default.nix
pkgs/development/compilers/gcc/4.7/default.nix
The 4.7 had some weird parameters added in crossAttrs; I've removed
them, but I don't understand where they come from.
This is for consistency with terminology in stdenv (and the terms
"hostDrv" and "buildDrv" are not very intuitive, even if they're
consistent with GNU terminology).
This allows various applications. It allows users to set global
optimisation flags, e.g.
stdenv.userHook = ''NIX_CFLAGS_COMPILE+=" -funroll-loops"'';
But the impetus is as an alternative to issue #229, allowing impure
stdenv setup for people who want to use distcc:
stdenv.userHook = "source /my/impure/setup-script.sh";
This is probably a bad idea, but at least now it's a bad idea in
people's configuration and not in Nixpkgs. :-)
This adapter causes the resulting binaries to have debug info and no
optimisations. Example use (in all-packages.nix):
foo = callPackage ./foo.nix {
stdenv = keepDebugInfo stdenv;
};
If the environment variable HYDRA_DISALLOW_UNFREE is set to "1", then
evaluation of a package with license "unfree" will throw an error.
Thus such packages or any packages that depend on them will fail to
evaluate.
