[Bug c/96165] New: internal compiler error during IPA pass: fnsummary

[jpalus+gcc at fastmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96165

Bug ID: 96165
   Summary: internal compiler error during IPA pass: fnsummary
   Product: gcc
   Version: 10.1.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: c
  Assignee: unassigned at gcc dot gnu.org
  Reporter: jpalus+gcc at fastmail dot com
  Target Milestone: ---

Created attachment 48863
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48863&action=edit
preprocessed source file

During Mesa compilation gcc segfaults with:


during IPA pass: fnsummary
../src/gallium/drivers/softpipe/sp_tex_sample.c: In function
‘sample_compare’:
../src/gallium/drivers/softpipe/sp_tex_sample.c:3862:1: internal compiler
error: Segmentation fault

Confirmed on both aarch64 and x86_64, note that issue occurs only if -O flag
was passed. Attached source file after preprocessor. Command to reproduce:

   gcc -O2 -c test.c -o test.o

[Bug target/104707] gcc on riscv64 isn't passing entire library path to the linker

[jpalus+gcc at fastmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104707

--- Comment #3 from Jan Palus  ---
I doubt this has anything to do with riscv or multilib. It's just gcc that
assumes all linkers should lookup /lib and /usr/lib and skips them when passing
library paths to linker:

https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/gcc.cc;h=bb07cc244e30fbeccc701816db888f497d65eb08;hb=refs/heads/master#l7931

I had the same issue on arm (32-bit) so no riscv, no multilib and dropping this
logic fixes the issue.

[Bug target/104707] gcc on riscv64 isn't passing entire library path to the linker

[jpalus+gcc at fastmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104707

--- Comment #5 from Jan Palus  ---
> Arm linux multilib is just as complex as riscvs really.

Perhaps my understanding of multilib is wrong but I meant pure 32bit, single
arch environment.

[Bug c/121044] New: False positive -Warray-bounds iwith GCC 14 and NEON intrinsics

[jpalus+gcc at fastmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121044

Bug ID: 121044
   Summary: False positive -Warray-bounds iwith GCC 14 and NEON
intrinsics
   Product: gcc
   Version: 14.3.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: c
  Assignee: unassigned at gcc dot gnu.org
  Reporter: jpalus+gcc at fastmail dot com
  Target Milestone: ---

Created attachment 61846
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=61846&action=edit
reproducer

Based on mbedtls build failure (upstream report
https://github.com/Mbed-TLS/mbedtls/issues/9003) attached reproducer for false
positive -Warray-bounds when targeting aarch64:

$ gcc -O2 -Warray-bounds repro.c -c -o repro.o 
In function ‘xor’,
inlined from ‘foo’ at repro.c:27:5:
repro.c:19:17: warning: array subscript 48 is outside array bounds of ‘unsigned
char[48]’ [-Warray-bounds=]
   19 | r[i] = a[i] ^ b[i];
  |~^~~
repro.c: In function ‘foo’:
repro.c:25:19: note: at offset 48 into object ‘tmp’ of size 48
   25 | unsigned char tmp[48] = { 0 };
  |   ^~~
In function ‘xor’,
inlined from ‘foo’ at repro.c:27:5:
repro.c:19:14: warning: array subscript 48 is outside array bounds of ‘unsigned
char[48]’ [-Warray-bounds=]
   19 | r[i] = a[i] ^ b[i];
  | ~^
repro.c: In function ‘foo’:
repro.c:25:19: note: at offset 48 into object ‘tmp’ of size 48
   25 | unsigned char tmp[48] = { 0 };
  |   ^~~

Note that it appears to affect GCC 14.x only (13.x/15.x seem to work fine
according to godbolt).

[Bug target/121044] [14 Regression] False positive -Warray-bounds iwith GCC 14 and NEON intrinsics

[jpalus+gcc at fastmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121044

--- Comment #2 from Jan Palus  ---
Some weird observations from reproducer minimazation:
- memcpy() needs to be there otherwise it's not reproduced (although it's
completely unrelated?)
- memcpy() needs to copy into array which is struct's member. plain array does
not trigger issue
- struct with array member needs to be passed as function argument

[Bug target/121044] [14 Regression] False positive -Warray-bounds with GCC 14 and vectors

[jpalus+gcc at fastmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121044

--- Comment #8 from Jan Palus  ---
Bisect results:

Regression between 13 and 14 started to appear with:

a291237b628f419d7f7ac264dd7b42947b565222 is the first bad commit
commit a291237b628f419d7f7ac264dd7b42947b565222
Author: Andrew MacLeod 
Date:   Tue Oct 31 16:51:34 2023

Remove simple ranges from trailing zero bitmasks.


Later fixed between 14 and 15 with:

74ee12ff68243bb177fb8653474dff80c3792139 is the first new commit
commit 74ee12ff68243bb177fb8653474dff80c3792139
Author: Andrew MacLeod 
Date:   Wed Jun 5 21:12:27 2024

Move array_bounds warnings into a separate pass.