[gcc-linaro wrong-code regression] gcc 4.8.1-2ubuntu1 to 4.8.1-3ubuntu1 breaks gtk on armhf

[Sebastien Bacher]

* Something in that upload is creating problems for gtk on armhf:
https://launchpad.net/ubuntu/+source/gcc-4.8/4.8.1-3ubuntu1

* The tests started failing, as showed by the saucy test rebuild:
https://launchpadlibrarian.net/142386450/buildlog_ubuntu-saucy-armhf.gtk%2B3.0_3.8.2-0ubuntu4_FAILEDTOBUILD.txt.gz

tests/a11y/text fails on a "Cannot animate property 'background-image'" error

The error can be reproduced on a porter box, downgrading gcc doing "make clean; make" in debian/build/shared/gtk leads to a working test. Building with -O0 workaround the problem, it fails when using -O1 or -O2 though

[Ktkachov (ktkachov)]

We're getting a miscompare on the 178.galgel benchmark in SPEC2000 on armv7l-unknown-linux-gnueabihf, on a Cortex-A15.
The .mis file just says 'galgel.out' short.

Bisection shows it started with r199439.

2013-05-30 Bernd Schmidt <email address hidden>
Zhenqiang Chen <email address hidden>

    config/arm/arm-protos.h: Add and update function protos.
    config/arm/arm.c (use_simple_return_p): New added.
    (thumb2_expand_return): Check simple_return flag.
    config/arm/arm.md: Add simple_return and conditional simple_return.
    config/arm/iterators.md: Add iterator for return and simple_return.

The compiler was configured: "--with-cpu=cortex-a15 --with-fpu=neon-vfpv4 --with-mode=thumb --with-float=hard --enable-languages=c,c++,fortran"

The flags for the benchmark are -ffixed-form -mthumb -mcpu=cortex-a15 -mfpu=neon-vfpv4 -mfloat-abi=hard -O3.
It works with -O1, fails at -O2 and -O3

[Zhenqiang Chen (zhenqiang-chen)]

I had reproduced it on chrome book. It failed due to "alignment exception" for O3.

I will do more investigation.

If this bug blocks your work, please try "-fno-shrink-wrap" to disable the optimization.

[Ktkachov (ktkachov)]

Thanks, Zhenqiang.

For me, it's a segfault in function DGETF2. gdb backtrace:
Program received signal SIGSEGV, Segmentation fault.
0x0001d964 in dgetf2_ ()
(gdb) bt
#0 0x0001d964 in dgetf2_ ()
#1 0x000255a4 in dgetrf_ ()
#2 0x00015856 in sysnsl_ ()
#3 0x0001194e in MAIN__ () at galgel.f90:9

Looking further into it, the segfault seems to be from the function idamax that is inlined in dgetf2.

Hope this helps,
Kyrill

[Zhenqiang Chen (zhenqiang-chen)]

Created attachment 30330
pr57637.patch

[Zhenqiang Chen (zhenqiang-chen)]

Root cause:

For ARM, gcc might generate cond_exec insns before shrink-wrapping. But DF_LR_BB_INFO (bb)->def does not include def info from cond_exec insns. So the check in function move_insn_for_shrink_wrap is not enough.

We'd check DF_LIVE_BB_INFO (bb)->gen.

Kyrill,

Please try the attached patch. If it works, I will send it for community review.

Thanks!
-Zhenqiang

[Ktkachov (ktkachov)]

> Please try the attached patch. If it works, I will send it for community
> review.

Hi Zhenqiang,

This seems to fix the miscompare on my setup, thanks!
I haven't ran a full regression test on it though, but it seems reasonable at first glance.

Kyrill

[Mikpe (mikpe)]

(In reply to zhenqiang.chen from comment #3)
> Created attachment 30330 [details]
> pr57637.patch

This patch breaks Ada bootstrap on x86_64 for me:

/tmp/objdir/./prev-gcc/xgcc -B/tmp/objdir/./prev-gcc/ -B/tmp/install49/x86_64-unknown-linux-gnu/bin/ -B/tmp/install49/x86_64-unknown-linux-gnu/bin/ -B/tmp/install49/x86_64-unknown-linux-gnu/lib/ -isystem /tmp/install49/x86_64-unknown-linux-gnu/include -isystem /tmp/install49/x86_64-unknown-linux-gnu/sys-include -c -g -O2 -gtoggle -gnatpg -W -Wall -g -O1 -fno-inline \
 -nostdinc -I- -I. -Iada -I/tmp/gcc-4.9-20130616/gcc/ada -I/tmp/gcc-4.9-20130616/gcc/ada/gcc-interface /tmp/gcc-4.9-20130616/gcc/ada/a-except.adb -o ada/a-except.o

raised STORAGE_ERROR : stack overflow or erroneous memory access
make[3]: *** [ada/a-except.o] Error 1
make[3]: Leaving directory `/tmp/objdir/gcc'
make[2]: *** [all-stage2-gcc] Error 2
make[2]: Leaving directory `/tmp/objdir'
make[1]: *** [stage2-bubble] Error 2
make[1]: Leaving directory `/tmp/objdir'
make: *** [bootstrap] Error 2

[Matthias Klose (doko)]

did work with the -2ubuntu1 upload, and does work with a gcc-4.8 build with the linaro changes disabled

[Sebastien Bacher (seb128)]

(the issue is not specific to this test, the same warning is hit when using that libgtk-3.so binary on arm)

[Matthias Klose (doko)]

current test case is to build gtk3 in saucy, and run the gtk3 tests

[Sebastien Bacher (seb128)]

Some details on how I test atm:

- apt-get source gtk+3.0
- cd gtk+-3.0-3.8.2
- debuild

That leads to a failing build on test

if you want to just run one of the failing tests:
- xvfb-run debian/build/share/tests/a11y/text

To test changes (e.g downgrading gcc or -O0):
- cd debian/build/share/gtk
- make clean
- (edit Makefile to change -O)
- make
- xvfb-run ../tests/a11y/text

-> gives a working test.

I'm trying to figure out what option exactly is breaking it, I didn't find yet what objects in gtk need to be rebuilt to trigger the problem or not though...

[Sebastien Bacher (seb128)]

updated/smaller testcase:
- untar the gtk source
- autoreconf
- ./configure --prefix=/usr --disable-test-print-backend --enable-x11-backend --disable-gtk-doc --enable-shared --disable-introspection --disable-colord --disable-static --disable-wayland-backend
(turning off some of the options helps to lower the turnover)
- make
- cd gtk
- xvfb-run ./tests/builder

-> that breaks

- sed -i 's#-O2# #g' Makefile
- touch gtkcssstylepropertyimpl.c; make; xvfb-run ./tests/builder

-> works

[Zhenqiang Chen (zhenqiang-chen)]

Created attachment 30364
update patch

Please try the updated patch. Local tests: x86-64 and pandaboard bootstrap OK.

[Ktkachov (ktkachov)]

(In reply to zhenqiang.chen from comment #7)
> Created attachment 30364 [details]
> update patch
>
>
> Please try the updated patch. Local tests: x86-64 and pandaboard bootstrap

This fixes the segfault in SPEC on arm for me.

Is there a way to produce a testsuite testcase for this regression? The benchmark where this occurs is a large chunk of Fortran, not sure how easy it is to reduce it.

[Zhenqiang Chen (zhenqiang-chen)]

I have no environment to reproduce it.

Would you please try option "-fno-shrink-wrap"? If it works, the root cause maybe the same as http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57637. In this case, would you please try the update patch at pr57637? If it still fails (without -fno-shrink-wrap option), please add option "-fdump-rtl-pro_and_epilogue" and attach the gtkcssstylepropertyimpl.c.*.pro_and_epilogue.

Thanks!
-Zhenqiang

[Sebastien Bacher (seb128)]

> Would you please try option "-fno-shrink-wrap"

That fix it, thanks!

> would you please try the update patch at pr57637?

@doko: is there any chance you could include that one in the next upload or in a test package? I'm not sure I want to try rebuilding gcc on the panda board, and I don't have root access to replace the compiler with the test version...

[Rearnsha (rearnsha)]

Marked as confirmed based on subsequent discussion

[Zhenqiang Chen (zhenqiang-chen)]

I had confirmed with 2013.06 release configured by --enable-languages=c,c++ --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb on Chrome book.

The root cause is the same as pr57637.

The failed functions is gtk_css_style_property_register.

[Ktkachov (ktkachov)]

I can confirm as of r201326:

    2013-07-30 Zhenqiang Chen <email address hidden>

        PR rtl-optimization/57637
        * function.c (move_insn_for_shrink_wrap): Also check the
        GEN set of the LIVE problem for the liveness analysis
        if it exists, otherwise give up.

the segfault is gone.

Thanks,
Kyrill

[Matthias Klose (doko)]

will be fixed with new GCC upload, reopen the gtk task to remove the workaround after the GCC upload

[Matthias Klose (doko)]

fixed in saucy

[Hjl-0 (hjl-0)]

Author: hjl
Date: Mon Mar 31 16:21:30 2014
New Revision: 208963

URL: http://gcc.gnu.org/viewcvs?rev=208963&root=gcc&view=rev
Log:
Backport revision 201326

gcc/

 PR rtl-optimization/60700
 2013-07-30 Zhenqiang Chen <email address hidden>

 PR rtl-optimization/57637
 * function.c (move_insn_for_shrink_wrap): Also check the
 GEN set of the LIVE problem for the liveness analysis
 if it exists, otherwise give up.

gcc/testsuite/

 PR rtl-optimization/60700
 2013-07-30 Zhenqiang Chen <email address hidden>

 * gcc.target/arm/pr57637.c: New testcase.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.target/arm/pr57637.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/function.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog