Pre-commit execution test for pr113363.f90
Hi there, Thanks for the heads-up on gcc patch #88281: 6 regressions on arm. I see from the log that the test timed-out and the core was dumped. I cannot reproduce this and can see nothing in the tree-dump that might cause a time out. I would appreciate some help on where the fault lies and what the cause might be. Best regards Paul Thomas ___ linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org
Re: Pre-commit execution test for pr113363.f90
[CC: Thiago, for GDB crash] Hi Paul, The test crashes immediately, it doesn't time out. You see "timeout" in the output because we run all tests under "timeout" utility. The backtrace is from the crash is: === rogram received signal SIGSEGV, Segmentation fault. 0xf7da9070 in arena_for_chunk (ptr=0x26208) at arena.c:156 156 arena.c: No such file or directory. (gdb) bt #0 0xf7da9070 in arena_for_chunk (ptr=0x26208) at arena.c:156 #1 arena_for_chunk (ptr=0x26208) at arena.c:160 #2 __GI___libc_free (mem=) at malloc.c:3390 #3 0x000125e8 in p () at /home/maxim.kuvyrkov/tcwg_gnu/abe/snapshots/gcc.git~master/gcc/testsuite/gfortran.dg/pr113363.f90:49 === which is the line ... (gdb) up #3 0x000125e8 in p () at /home/maxim.kuvyrkov/tcwg_gnu/abe/snapshots/gcc.git~master/gcc/testsuite/gfortran.dg/pr113363.f90:49 49deallocate (x, y) (gdb) p y $1 = ( _data = 0x26238, _vptr = 0x24090 <__vtab_CHARACTER_1_.7>, _len = 10 ) However, if I try to print "x", GDB crashes: === (gdb) p x $2 = ( _data = (0x6568, /build/gdb-aPmCGS/gdb-12.1/gdb/value.c:856: internal-error: value_contents_bits_eq: Assertion `offset1 + length <= TYPE_LENGTH (val1->enclosing_type) * TARGET_CHAR_BIT' failed. A problem internal to GDB has been detected, further debugging may prove unreliable. - Backtrace - 0x911165 ??? 0xb69e3b ??? 0xb69fdd ??? 0xc9b135 ??? 0xb78c97 ??? 0x9d6d65 ??? 0x9d7513 ??? 0x9d6813 ??? 0xb74e23 ??? 0x9d66e7 ??? 0xb74e23 ??? 0x92b193 ??? 0xb74f5b ??? 0xa81f3f ??? 0xa820f9 ??? 0x935699 ??? 0xb41929 ??? 0x9c7023 ??? 0x9c72e9 ??? 0x9c7973 ??? - /build/gdb-aPmCGS/gdb-12.1/gdb/value.c:856: internal-error: value_contents_bits_eq: Assertion `offset1 + length <= TYPE_LENGTH (val1->enclosing_type) * TARGET_CHAR_BIT' failed. A problem internal to GDB has been detected, further debugging may prove unreliable. === So try looking at the dumps of how "x" is created/freed. The testcase is at https://people.linaro.org/~maxim.kuvyrkov/pr113363.exe ; it was compiled with: /home/maxim.kuvyrkov/tcwg_gnu/abe/builds/destdir/armv8l-unknown-linux-gnueabihf/bin/armv8l-unknown-linux-gnueabihf-gfortran /home/maxim.kuvyrkov/tcwg_gnu/abe/snapshots/gcc.git~master/gcc/testsuite/gfortran.dg/pr113363.f90 -fdiagnostics-plain-output -fdiagnostics-plain-output -g -pedantic-errors -L/home/maxim.kuvyrkov/tcwg_gnu/abe/builds/armv8l-unknown-linux-gnueabihf/armv8l-unknown-linux-gnueabihf/gcc-gcc.git~master-stage2/armv8l-unknown-linux-gnueabihf/./libgfortran/.libs -L/home/maxim.kuvyrkov/tcwg_gnu/abe/builds/armv8l-unknown-linux-gnueabihf/armv8l-unknown-linux-gnueabihf/gcc-gcc.git~master-stage2/armv8l-unknown-linux-gnueabihf/./libatomic/.libs -lm -o ./pr113363.exe It should run on any stock ubuntu-22.04 rootfs for armhf architecture. Kind regards, -- Maxim Kuvyrkov https://www.linaro.org > On Apr 10, 2024, at 18:31, Paul Richard Thomas > wrote: > > Hi there, > > Thanks for the heads-up on gcc patch #88281: 6 regressions on arm. > > I see from the log that the test timed-out and the core was dumped. I > cannot reproduce this and can see nothing in the tree-dump that might cause > a time out. I would appreciate some help on where the fault lies and what > the cause might be. > > Best regards > > Paul Thomas > ___ > linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org > To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org ___ linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org
Re: Pre-commit execution test for pr113363.f90
Maxim Kuvyrkov writes: > [CC: Thiago, for GDB crash] Thanks! > However, if I try to print "x", GDB crashes: > === > (gdb) p x > $2 = ( _data = (0x6568, > /build/gdb-aPmCGS/gdb-12.1/gdb/value.c:856: internal-error: > value_contents_bits_eq: > Assertion `offset1 + length <= TYPE_LENGTH (val1->enclosing_type) * > TARGET_CHAR_BIT' > failed. > A problem internal to GDB has been detected, > further debugging may prove unreliable. I can reproduce this crash with GDB 12.1 from Ubuntu 22.04, but testing with current tip of trunk detects the corruption in the inferior: Program received signal SIGSEGV, Segmentation fault. 0xf7da9070 in arena_for_chunk (ptr=0x26208) at arena.c:156 warning: 156arena.c: No such file or directory (gdb) bt #0 0xf7da9070 in arena_for_chunk (ptr=0x26208) at arena.c:156 #1 arena_for_chunk (ptr=0x26208) at arena.c:160 #2 __GI___libc_free (mem=) at malloc.c:3390 #3 0x000125e8 in p () at /home/thiago.bauermann/src/gcc/gcc/testsuite/gfortran.dg/pr113363.f90:49 (gdb) frame 3 #3 0x000125e8 in p () at /home/thiago.bauermann/src/gcc/gcc/testsuite/gfortran.dg/pr113363.f90:49 49deallocate (x, y) (gdb) p x $1 = ( _data = (, _vptr = 0x24090 <__vtab_CHARACTER_1_.7>, _len = 6 ) (gdb) GDB 14.1 was the first version that didn't crash, but it shows garbage contents for x._data: (gdb) p x $1 = ( _data = (0x6568, 0x65), _vptr = 0x24090 <__vtab_CHARACTER_1_.7>, _len = 6 ) GDB 15.1 will be the first version that shows the out of bounds error message. -- Thiago ___ linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org
Re: [Linaro-TCWG-CI] v6.9-rc2-36-gfc4216add64ee: Failure on arm
ci_not...@linaro.org writes: > In CI config tcwg_kernel/gnu-master-arm-next-allyesconfig after: > > | commit v6.9-rc2-36-gfc4216add64ee > | Author: Samuel Holland > | Date: Fri Mar 29 00:18:28 2024 -0700 > | > | drm/amd/display: use ARCH_HAS_KERNEL_FPU_SUPPORT > | > | Now that all previously-supported architectures select > | ARCH_HAS_KERNEL_FPU_SUPPORT, this code can depend on that symbol > instead > | of the existing list of architectures. It can also take advantage of > the > | common kernel-mode FPU API and method of adjusting CFLAGS. > | > | ... 21 lines of the commit log omitted. > > Results changed to > # reset_artifacts: > -10 > # build_abe binutils: > -9 > # build_abe stage1: > -5 > # build_abe qemu: > -2 > # linux_n_obj: > 22511 > > From > # reset_artifacts: > -10 > # build_abe binutils: > -9 > # build_abe stage1: > -5 > # build_abe qemu: > -2 > # linux_n_obj: > 22321 > # linux build successful: > all This problem still reproduces on next-20240410, so I reported it to the patch author. He responded here: https://lore.kernel.org/linux-arm-kernel/75a37a4b-f516-40a3-b6b5-4aa1636f9...@sifive.com/ mentioning that it's a shortcoming of GCC on arm: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91970 Though Andrew Pinski has just disagreed with that assessment. -- Thiago ___ linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org
