[lfs-dev] Recent patches

Ken Moffat zarniwhoop at ntlworld.com
Tue Feb 16 20:16:20 PST 2016


On Tue, Feb 16, 2016 at 07:48:48PM -0600, Bruce Dubbs wrote:
> I've gone ahead and committed changes to incorporate patches for recent
> upstream commits.  After a full build, I'm a little uncomfortable about the
> gcc patch.  It is 7648 lines and the build shows several unexpected failures
> that were not present before.
> 
> Running target unix
> FAIL: 20_util/shared_ptr/cons/58659.cc (test for excess errors)
> UNRESOLVED: 20_util/shared_ptr/cons/58659.cc compilation failed to produce
> executable

gcc bug 58659 was about 18 months ago, so that looks like a
regression

https://gcc.gnu.org/ml/gcc-bugs/2013-10/msg00451.html

> FAIL: experimental/filesystem/iterators/directory_iterator.cc execution test
> FAIL: experimental/filesystem/iterators/recursive_directory_iterator.cc
> execution test
> 

for experimental - on its own - I worry less, but if the patch
introduces it (I didn't look to see if the test is new, too busy
rebuilding systems) it doesn't look good.

>       === libstdc++ Summary ===
> 
> # of expected passes    10365
> # of unexpected failures   3

[...]
> 
> Running target unix
> FAIL: gcc.dg/torture/pr29119.c   -O0  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -O1  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -O2  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -O3 -fomit-frame-pointer  (test for excess
> errors)
> FAIL: gcc.dg/torture/pr29119.c   -O3 -g  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -Os  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -O2 -flto -fno-use-linker-plugin
> -flto-partition=none  (test for excess errors)
> FAIL: gcc.dg/torture/pr29119.c   -O2 -flto -fuse-linker-plugin
> -fno-fat-lto-objects  (test for excess errors)

On some of my boxes, at verious times, failures in the torture tests
have been common.  But if the patch introduced it then again it is a
regression.  29119 was over 9 years ago -

https://gcc.gnu.org/ml/gcc-bugs/2006-09/msg01638.html

> FAIL: gcc.target/i386/avx512dq-vextractf64x2-1.c scan-assembler-times
> vextractf64x2[ \\\\t]+[^{\\n]*%ymm[0-9]+.{7}(?:\\n|[ \\\\t]+#) 1

Might be a new test.  But overall, these look like regressions.
> 
> How should we approach this?  The easy way is to not apply the new patch.
> Additionally I do not know what the patch is supposed to fix.
> 
> How should we go on this?  I will hold off on -rc2 until we can reach a
> consensus.
> 
I'm late, but I agree with not applying the patch at this stage.

ĸen
-- 
This email was written using 100% recycled letters.


More information about the lfs-dev mailing list