the target libraries, as 4.3.4
I trigger the breaking build using uclibc, instead of glibc, and then the
mixture of CPP getting the build glibc headers, and CC getting the uclibc
headers, makes the problem evident.
In 4.3.4 it only affected libmudflap. In 4.4.2, it affects libmudflap and
libstdc++.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18850
through libtool.
I found a make variable holding flags that will not be trimmed by libtool:
FLAGS_FOR_TARGET.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18846
Dealt with the gcc 4.4 expression for it to work well for a native and a cross
compiler, adding a pair of patches to get the job done without much side
effects. One (libtool problems on -B) comes from the gcc 4.3 expression, while
the other fixes a bug already fixed in gcc 4.4.3.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18833
between uclibc/glibc still.
I started the renaming from glibc to libc regarding the cross-toolchain, but I
still have to finish.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18801
I also set the 'glibcLocales' top-level attribute point to 2.11 instead of
2.10, to match that of the 'glibc' top-level attribute.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18746
used to cross-compile, because of some bug in 4.4.2 I think. This bug does not
allow building a cross-compiler without libc:
http://archives.free.net.ph/message/20091119.140407.040d3c5b.en.html
Maybe we should wait for 4.4.3 or stay with 4.4.1 cross compiling. By now I
will keep 4.3.4 on cross compilation.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18730
Fixing the glibc-2.10 expression on cross-builds (which should be ported to
the glibc-2.11 expression once we get "ports" there)
Making kde3 and cyrus-sasl use gcc-4.3, because the strictness in gcc-4.4 does
not allow them build.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18706
- I still have not understood why it worked without this fix before, and I think
this has been triggered by the gcc-4.4, but I have not investigated this much. I
went with the trivial fix.
Adding a glibc-2.10.1 expression, because the glibc-2.11 still does not have
a ports release, so it cannot be used in arm. I'm using it only in native
compilation by now.
Making the default glibc to be 2.10 instead of 2.11 in armv5tel-linux.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18688
renaming.
I think directory renaming breaks the usual merges... because it leaves the
'to be removed' directory in the working directory still. A manual 'rm' of the
'to be removed' directory fixed the commit.
svn merge ^/nixpkgs/trunk
svn path=/nixpkgs/branches/stdenv-updates/; revision=18661
native strip. So we now distinguish dontStrip and dontCrossStrip. I updated
the expressions for glibc-2.9 and glibc-2.11 accordingly.
I could get rid of the cross-glibc depending on the cross-gcc-stage-static.
Enabling nls in the final cross-gcc.
I still have problems on wint_t/wchar_t not working on cross build. Gettext
does not build.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18562
- Disabling guile test, because one fails. I commented on that in the source.
On cross builds:
- Adding stripping
- Updating the glibc-2.11 expression to match the parameters of glibc-2.9,
which I was updating more.
- Renaming from selfNativeBuildInput to selfBuildNativeInput, so this matches
better the pattern buildNativeInputs.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18550
- Before this changes, cflags and ldflags for the native and the cross compiler
got mixed. Not all the gcc-wrapper/gcc-cross-wrapper variables are
independant now, but enough, I think.
- Fixed the generic stdenv expression, which did a big mess on buildInputs and
buildNativeInputs. Now it distinguishes when there is a stdenvCross or not.
Maybe we should have a single stdenv and forget about the stdenvCross
adapter - this could end in a stdenv a bit complex, but simpler than the
generic stdenv + adapter.
- Added basic support in pkgconfig for cross-builds: a single PKG_CONFIG_PATH
now works for both the cross and the native compilers, but I think this
should work well for most cases I can think of.
- I tried to fix the guile expression to cross-biuld; guile is built, but not
its manual, so the derivation still fails. Guile requires patching to
cross-build, as far as I understnad.
- Made the glibcCross build to be done through the usage of a
gcc-cross-wrapper over the gcc-cross-stage-static, instead of using it
directly.
- Trying to make physfs (a neverball dependency) cross build.
- Updated the gcc expression to support building a cross compiler without getting
derivation variables mixed with those of the stdenvCross.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18534
I was trying to cross compile SDL. Many dependencies work, but I ended seeing
libX11 not ready for cross compilation. Other xorg libraries cross-compile
well. libX11 may need a small patch. The problem is the usual "configure test
cannot be run in cross compilation", so the configure script halts.
I made the pkgconfig expression always return buildDrv, as I think it rarely
will be needed as buildInput. So to avoid rewriting all its mentions to use
it as buildNativeInput, I prefered this small change.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18500
- Stating better the guile dependencies (native/host) for guile to build
- Fixing cross-linking, through --rpath-link (ld(1) explains well about it
- Made gcc call the linker and the assembler through the gcc wrapper instead of
directly. I thought this was the source of missing -rpath's, but the source
of the problem ended up being the lack of --rpath-link. But I think the
native gcc calls the wrapped ld and as, so let's do the same cross
compiling.
- Removed the binutilsCross from the glibc expressions. Now they are built
using the gcc-cross-wrapper, and they were built with the direct gcc and
binutils before this change.
- I think patchelf and strip don't break the cross-compiled binaries, so I
reallow them on cross compilation.
- I disable the checkPhase on cross compilation. This made gmp and libtool
fail when cross compiled, iirc.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18498
linking path), and with this achieved bash being cross-compilable.
I fixed the few expressions involved in bash building, so they have well stated
native and non-native inputs.
I also tried to cross-build guile, and with this I found a problem in the
actual cross-gcc: it calls the binutils ld, instead of the ld wrapper. This
way, the programs/shared_libraries don't get the proper -rpath.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18497
true/false, which tells whether the derivation needs itself as
buildNativeInput.
For example, in order to build cross ncurses, we need the a native build
ncurses.
(As libtool does not work in stdenv, I have not tested this change, to check
whether finally ncurses cross-build)
svn path=/nixpkgs/branches/stdenv-updates/; revision=18489
arguments for the ncurses expression.
We should find a way to express a dependency in cross compilation of the style
"cross-ncurses depends on having the native-ncurses".
svn path=/nixpkgs/branches/stdenv-updates/; revision=18479
all the naming in nixpkgs to match the new build/host cross compilation stdenv.
Nevertheless, we decided not to do the renaming, but I forgot this change in
readline until ludo told me about it.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18474
the cross compilation functionality.
- I renamed some expected stdenv.mkDerivation parameter attributes so we can
keep this branch properly updated from trunk. We agreed with Nicolas Pierron
doing a massive renaming, so all current buildInputs become hostInputs (input
as build for the host machine, in autotools terminology) , and
then buildInputs would mean "input as for the build machine".
By now, the specific "input as for the build machine" is specified through
buildNativeInputs. We should fix this in the merge to trunk.
- I made the generic stdenv understand the buildNativeInputs, otherwise if
we start changing nixpkgs expressions so they distinguish the current
buildInputs into buildInputs and buildNativeInputs, we could break even more
nixpkgs for other platforms.
- I changed the default result of mkDerivation so it becomes the derivation for
to be run in the build machine. This allows, without any special rewriting,
"fetchurl" derivations to be always results for the build machine to use
them.
- The change above implies that, for anyone wanting to cross-compile, has to
build the hostDrv of the wanted derivation. For example, after this commit,
the usual test of "nix-build -A bison.hostDrv arm.nix" works. I described
the contents of this arm.nix in r18398.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18471
`selectVersion ./foo "bar"' instead of `import ./foo/bar.nix'.
* Replaced `with args' with formal function arguments in several
packages.
* Renamed several files to `default.nix'. As a general rule, version
numbers should only be included in the filename when there is a
reason to keep multiple versions of a package in Nixpkgs.
Otherwise, it just makes it harder to update the package.
svn path=/nixpkgs/trunk/; revision=18403
needing to keep a new tree of expressions apart for the expressions to get
cross-compiled.
I changed the whole way of using cross compilation with nixpkgs, which before
was done through a simple adapter.
Now the adapter became complex, and I've tried to avoid the most obvious
recursivities. For example, the fetchurl expression should
never be cross-compiled, as the gmp, mpfr, and some others, like
some ncurses, perl, ... I made overrided copies of those necessary as
perlNoCross, ncursesNoCross, as stdenvNoCross, keeping in mind that
the stdenv (capable of cross compilation) is built upon stdenvNoCross using
an adapter.
So, to cross compile, instead of building using "nixpkgs/default.nix",
you should build with your
own "myarchiteture.nix", which should have contents like these, for example:
import /etc/nixos/nixpkgs/default.nix
{
crossSystem = {
config = "armv5tel-unknown-linux-gnueabi";
bigEndian = false;
arch = "arm";
float = "soft";
};
}
svn path=/nixpkgs/branches/stdenv-updates/; revision=18398
It still does not work, but I think I already get glibc cross compiled.
Next: gcc and g++, and set some setup script hooks on stdenvCross.
It took quite enough hours for this commit.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18351
My idea is to provide special stdenv expressions that will contain in the path
additional cross compilers. As most expressions for programs accept a stdenv parameter,
we could substitute this parameter with the special stdenv, which will have a
generic builder that attempts the usual "--target=..." and can additionally
have an env variable like "cross" with the target architecture set.
So, finally we could have additional expressions like this:
bashRealArm = makeOverridable (import ../shells/bash) {
inherit fetchurl bison;
stdenv = stdenvCross "armv5tel-unknown-linux-gnueabi";
};
Meanwhile it does not work - I still cannot get the cross-gcc to build.
I think it does not fill the previous expressions with a lot of noise, so I
think it may be a good path to follow.
I only touched some files of the current stdenv: gcc-4.3, kernel headers
2.6.28, glibc 2.9, ...
I tried to use the gcc-cross-wrapper, that may be very outdated. Maybe I will
update it, or update the gcc-wrapper expression to make it fit the cross tools,
but meanwhile I even cannot build gcc, so I have not tested the wrapper.
This new idea on cross compiling is not similar to that of the
nixpkgs/branches/cross-compilation, which mostly added bare new expressions for
anything to be cross compiled, if I understood it correctly.
I cared not to break anything of the usual stdenv in all this work.
svn path=/nixpkgs/branches/stdenv-updates/; revision=18343
This comes from:
svn diff ^/nixpkgs/trunk/@18255 ^/nixpkgs/branches/stdenv-updates/ > diff
patch -p0 < diff
and then adding into svn all files new from the patch.
trunk@18255 comes from the last time I updated stdenv-updates from trunk.
svn path=/nixpkgs/stdenv-updates2/; revision=18272
Redland's use of the pre-processor symbol SQLITE_API collides with a
symbol of the same name in sqlite 3.6.19. Patching the source code to
use REDLAND_SQLITE_API instead works around the problem.
svn path=/nixpkgs/trunk/; revision=18100
Sqlite has a build mode called "amalgamation" that gathers all 90+ source code
files into a single sqlite3.c file before compiling the library. Building
sqlite this way reportedly gives a 5-10% performance gain because the compiler
can perform more sophisticated optimizations.
svn path=/nixpkgs/trunk/; revision=18092
The new version requires Tcl at build time in order to construct sqlite3.h,
even when the --without-tcl option is passed to configure. That feels odd
because the sqlite web site does advertise "no dependencies", but then it's
probably not a big deal either. The build of the Tcl plugin for sqlite is still
disabled, so there is no run-time dependency.
svn path=/nixpkgs/trunk/; revision=18091
propagatedBuildInputs, because those inputs are required by the *.pc
or *.la files of the package:
- If a *.pc file references a non-propagated input, then Gnome
packages have the bad tendency to silently ignore this problem in
configure scripts - the failure of a command like `pkg-config
--cflags foo' will be ignored if a dependency of foo.pc is
missing, so no flags will be added, and the build will fail later
on a missing header or library.
- If a *.la file references a non-propagated input, the build will
also fail, because Libtool will add library dependencies that it
cannot find. (Arguably *.la files should never reference packages
that aren't in the corresponding *.pc file, but they do it
anyway).
By setting the propagatedBuildInputs properly, it should be possible
to get rid of all the NIX_CFLAGS_COMPILE / NIX_LDFLAGS hacks in the
Gnome expressions.
svn path=/nixpkgs/branches/xorg-7.5/; revision=18084
configuration directory so that users on other distributions don't
need to set $FONTCONFIG_FILE (NIXPKGS-29). Also use
/var/cache/fontconfig for the cache to prevent programs run by root
from writing into the Nix store.
svn path=/nixpkgs/branches/xorg-7.5/; revision=18021
patent-encumbered hinting and sub-pixel rendering. It's disabled by
default. (Or should it be enabled by default?)
svn path=/nixpkgs/branches/xorg-7.5/; revision=18019
development/libraries/{glib,gtk+,pango,atk,...}. Done for glib/gtk+
1.2. Also deleted some obsolete, unused versions (gtkLibs 2.10,
2.12, and 2.14).
svn path=/nixpkgs/trunk/; revision=17992
* Dropped "nolongdouble.patch". The patch no longer applies to Python 2.6, and
apparently isn't required anymore either.
* Added access to native Darwin arch utility. Python tries to run 'arch' in
the configure stage, but that binary reside in /usr/bin. To make it
available to the expression, the small wrapper darwinArchUtility is added as
a buildInput if appropriate.
* Don't pass --enable-shared. The build fails if we try to enable building of
shared libraries, apparently because some required libraries aren't linked,
i.e. the linker call isn't right.
TODO:
* Figure out how to enable shared linking.
* The resulting binary on Darwin seem to lack the binascii module.
svn path=/nixpkgs/trunk/; revision=17894
The build succeeds on i686-linux. Other platforms look good, too,
because there were hardly any changes necessary to update the expression
from 2.5.
svn path=/nixpkgs/trunk/; revision=17889
Updating libgpod
Making gtkpod accept 'ogg' files, and made it convert them well to mp3, if 'lame'
and oggdec is in path. It should better reference lame and libvorbis store path
files.
svn path=/nixpkgs/trunk/; revision=17888
The initial MacOS X binaries have been linked to libgmp.dylib using some
mad path in /opt that's now hard-coded into the program. Consequently,
$DYLD_LIBRARY_PATH must contain the place where libgmp really is for
those binaries to run correctly. Tested on i386-apple-darwin9.7.0.
svn path=/nixpkgs/trunk/; revision=17873
On MacOS X, we used to use the native perl interpreter from /usr/bin.
Unfortunately, that interpreter fails to build a number of packages
(Subversion, Git, etc. ...), because it assumes knowledge about the
underlying C compiler that is not valid for the compiler used by Nix.
For example, /usr/bin/perl assumes that the compiler can build binaries
for both the ppc and the x86 architecture. /usr/bin/gcc can do that, but
the gcc from Nix can't.
The solution is to compile Perl 5.10 in Nix so that the ./configure
phase can properly detect the system's capabilities. However, note that
the resulting binary is impure: it will find headers in /usr/include and
libraries in /usr/lib. In this respect, the Nix-compiled perl binary is
no different than the native one in /usr/bin -- it's just configured
more accurately.
svn path=/nixpkgs/trunk/; revision=17870
so all hooks on PERL5LIB include this module properly.
I don't know why this doesn't install the files to that */site_perl/* by default.
svn path=/nixpkgs/trunk/; revision=17837
