I made a mistake merge. Reverting it in c778945806 undid the state
on master, but now I realize it crippled the git merge mechanism.
As the merge contained a mix of commits from `master..staging-next`
and other commits from `staging-next..staging`, it got the
`staging-next` branch into a state that was difficult to recover.
I reconstructed the "desired" state of staging-next tree by:
- checking out the last commit of the problematic range: 4effe769e2
- `git rebase -i --preserve-merges a8a018ddc0` - dropping the mistaken
merge commit and its revert from that range (while keeping
reapplication from 4effe769e2)
- merging the last unaffected staging-next commit (803ca85c20)
- fortunately no other commits have been pushed to staging-next yet
- applying a diff on staging-next to get it into that state
Turns out that libgccjit gets installed as a .so file, which the gcc
builder.sh didn't change: It only touched .dylib files; that means
that anything linking in libgccjit.so would receive an "Image not
found" error at load time.
With this change, we invoke `install_name_tool` on .so files too,
adjusting their dynamic linker ID, so that they too can be found.
In some cases, such as when building cross compilers, the binaries and
manpages contain the target architecture tuple, such as
`i686-w64-mingw32-g++.1`.
Ensure the symlink created to save space with the duplicated manpage
(`g++.1 -> gcc.1`) properly handles such cases and generates symlinks
such as `i686-w64-mingw32-g++.1 -> i686-w64-mingw32-gcc.1`.
Previously in such cases, a broken `gcc.1` link would be created
instead.
I hate the thing too even though I made it, and rather just get rid of
it. But we can't do that yet. In the meantime, this brings us more
inline with autoconf and will make it slightly easier for me to write a
pkg-config wrapper, which we need.
libgccjit is a library but is used as a compiler. So it references a
bunch of compiler things in $out. To avoid a cycle, we need to put
everything in $out, so referenced to $lib need to be replaced with
${!outputLib}.
- NIX_CC_CROSS is now completely gone!
- NIX_CC is defined reliably, so no manual def needed
- stdenv.ccCross -> stdenv.cc, also removing need for "or" fallback
- disable bootstrap builds on Darwin
- remove xcrun calls
- check if patchelf is available before using
- apply darwin patch for gcc4.9
- fixes#16047
- fixes#14812
... because cc-wrapper is meant to propagate man pages into user envs,
and info pages are rather large.
Also replace the duplicate g++ and gcc man1 pages by a symlink.
- $out/libexec/ now contains subdirectories that we want to patchelf too
- $out/libexec/ now contains some .so and a related .la file, which must
not be passed to patchelf
be passed to derivations that need to apply patches.
* GCC 3.4 is now the default compiler (old GCC renamed to `gcc-3.3').
* The temporary GCCs built during the stdenvLinux bootstrap are now
built without C++ support and without profiling.
* Remove fixincl in GCC 3.4 to prevent a retained dependency on the
previous GCC.
* Always set $prefix in setup.sh, even when there is no configure
script.
svn path=/nixpkgs/trunk/; revision=1444
libgcc of the gcc being built, not the gcc building it.
* Only include a directory in the rpath of an executable/library if it
is actually used. Before, the `/lib' directory of every build input
was added to the rpath, causing many unnecessary retained
dependencies. For instance, Perl has a `/lib' directory, but most
applications whose build process uses Perl don't actually link
against Perl. (Also added a test for this.)
* After building glibc, remove glibcbug, to prevent a retained
dependency on gcc.
* Add a newline after `building X' in GNU Make.
svn path=/nixpkgs/trunk/; revision=911
* Make builders unexecutable by removing the hash-bang line and
execute permission.
* Convert calls to `derivation' to `mkDerivation'.
* Remove `system' and `stdenv' attributes from calls to
`mkDerivation'. These transformations were all done automatically,
so it is quite possible I broke stuff.
* Put the `mkDerivation' function in stdenv/generic.
svn path=/nixpkgs/trunk/; revision=874
whether the system header file directory actually exists (when
calling fixinc), so passing a non-existent directory no longer
works. Instead we make a empty dummy directory.
svn path=/nixpkgs/trunk/; revision=858
builders for typical Autoconf-style to be much shorten, e.g.,
. $stdenv/setup
genericBuild
The generic builder does lots of stuff automatically:
- Unpacks source archives specified by $src or $srcs (it knows about
gzip, bzip2, tar, zip, and unpacked source trees).
- Determines the source tree.
- Applies patches specified by $patches.
- Fixes libtool not to search for libraries in /lib etc.
- Runs `configure'.
- Runs `make'.
- Runs `make install'.
- Strips debug information from static libraries.
- Writes nested log information (in the format accepted by
`log2xml').
There are also lots of hooks and variables to customise the generic
builder. See `stdenv/generic/docs.txt'.
* Adapted the base packages (i.e., the ones used by stdenv) to use the
generic builder.
* We now use `curl' instead of `wget' to download files in `fetchurl'.
* Neither `curl' nor `wget' are part of stdenv. We shouldn't
encourage people to download stuff in builders (impure!).
* Updated some packages.
* `buildinputs' is now `buildInputs' (but the old name also works).
* `findInputs' in the setup script now prevents inputs from being
processed multiple times (which could happen, e.g., if an input was
a propagated input of several other inputs; this caused the size
variables like $PATH to blow up exponentially in the worst case).
* Patched GNU Make to write nested log information in the format
accepted by `log2xml'. Also, prior to writing the build command,
Make now writes a line `building X' to indicate what is being
built. This is unfortunately often obscured by the gigantic tool
invocations in many Makefiles. The actual build commands are marked
`unimportant' so that they don't clutter pages generated by
`log2html'.
svn path=/nixpkgs/trunk/; revision=845