proposing {gcc,gcj,gnat}-4.3 4.3.2-1 for testing
now that the packages are all built on all architectures I'm proposing
these packages for testing. Comparing the testsuite results from the
gcc-4.3 and gcj-4.3 4.3.1-9 and 4.3.2-1 packages there are no
regressions on the release architectures. gnat-4.3 currently doesn't
run the testsuite.
- gcc i386: gcc.target/i386/local2.c is a progression, reducing the
number of unexpected test failures to 0.
- sparc -m64: gcc.c-torture/unsorted/SFset.c is a progression.
- gcj armel: eight more passing tests
- gcj m68k: regression: Thread_Alive -O3
- there are some test timeouts in the libgomp testsuites on many
architectures and in the gfortran testsuite on hppa. These vary
depending on the buildd.
The upstream changes are regression fixes only, except:
- spu updates: checked these on a ps3 rebuilding libspe2 and running
the tests/examples.
- version bump. Note that all files are still located in the same
directories, and both the 4.3.1-9 and 4.3.2-1 uploads have both
4.3.x symlinks in the gcc_lib_dir.
gcj-4.3 4.3.2-2 fixes one more bug in gjavah causing bootstrap
failures for newer GCC versions.
I don't plan any further updates for these packages for now.
If all there packages are unblocked, please consider unblocking
gcc-defaults as well, adjusting just the package versions and
constraining the dependencies.
Matthias
changes in the packages:
gcc-4.3 (4.3.2-1) unstable; urgency=medium
[Matthias Klose]
* Final gcc-4.3.2 release (regression fixes).
- Remove the generated install docs from the tarball (GFDL licensed).
- C++ regression fixes: PR debug/37156.
- general regression fixes: PR debug/37156, PR target/37101.
- Java regression fixes: PR libgcj/8995.
* Update to SVN 20080905 from the gcc-4_3-branch.
- C++ regression fixes: PR c++/36741 (wrong diagnostic),
- general regression fixes: PR target/37184 (ice on valid code),
PR target/37191 (ice on valid code), PR target/37197 (ice on valid code),
PR middle-end/36817 (ice on valid code), PR middle-end/36548 (wrong code),
PR middle-end/37125 (wrong code), PR c/37261 (wrong diagnostic),
PR target/37168 (ice on valid code), PR middle-end/36449 (wrong code),
PR middle-end/37248 (missed optimization), PR target/36332 (wrong code).
- Fortran regression fixes: PR fortran/37193 (rejects valid code).
* Move symlinks in gcc_lib_dir from cpp-4.3 to gcc-4.3-base. Closes: #497369.
* Don't build-depend on autogen on architectures where it is not installable
(needed for the fixincludes testsuite only); don't build-depend on it for
source packages not running the fixincludes testsuite.
[Ludovic Brenta]
* Add sdefault.ads to libgnatprj4.3-dev. Fixes: #492866.
* turn gnatvsn.gpr and gnatprj.gpr into proper library project files.
* Unconditionally build-depend on gnat when building gnat-4.3.
Fixes: #487564.
* (debian/rules.d/binary-ada.mk): Add a symlink libgnat.so to
/usr/lib/libgnat-4.3.so in the adalib directory. Fixes: #493814.
* (debian/patches/ada-sjlj.dpatch): remove dangling symlinks from all
adalib directories.
* debian/patches/ada-alpha.dpatch: remove, applied upstream.
[Samuel Tardieu, Ludovic Brenta]
* debian/patches/pr16086.dpatch: new; backport from GCC 4.4.
Closes: #248172.
* debian/patches/pr35792.dpatch: new; backport from GCC 4.4.
* debian/patches/pr15808.dpatch (fixes: #246392),
debian/patches/pr30827.dpatch: new; backport from the trunk.
gcj-4.3 (4.3.2-2) unstable; urgency=low
* Fix naming of bridge targets in gjavah (wrong header generation).
Reply to: