https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103872
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target|riscv64-unknown-elf |riscv64-unknown-elf
| |hppa*-*-* aarch64-*-*
Status|UNCONFIRMED |NEW
Last reconfirmed| |2022-01-28
Ever confirmed|0 |1
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
strict alignment is causing the memcpy to have slightly different IR between
the targets.
I can reproduce the failure on aarch64 with -mstrict-align.
ubuntu@ubuntu:~/src/upstream-gcc-aarch64/gcc/objdir/gcc\# ./xgcc -B. t.c
-fanalyzer -Wanalyzer-too-complex -fanalyzer-call-summaries
ubuntu@ubuntu:~/src/upstream-gcc-aarch64/gcc/objdir/gcc\# ./xgcc -B. t.c
-fanalyzer -Wanalyzer-too-complex -fanalyzer-call-summaries -mstrict-align
t.c: In function 'game_new':
t.c:31:1: warning: leak of 'tmp.word_state' [CWE-401] [-Wanalyzer-malloc-leak]
31 | } /* { dg-bogus "leak" } */
| ^
'game_new': events 1-7
|
| 20 | if ((tmp.word_state = malloc(wordlen+1)) == NULL)
| | ~ ^~~~~~~~~~~~~~~~~
| | | |
| | | (1) allocated here
| | (2) assuming 'tmp.word_state' is non-NULL
| | (3) following 'false' branch...
| 21 | goto err;
| 22 | if ((rval = malloc(sizeof(*rval))) == NULL)
| | ~ ~~~~~~~~~~~~~~~~~~~~~
| | | |
| | | (4) ...to here
| | (5) following 'false' branch (when 'rval' is
non-NULL)...
| 23 | goto err;
| 24 | memcpy(rval, &tmp, sizeof(*rval));
| | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| | |
| | (6) ...to here
|......
| 31 | } /* { dg-bogus "leak" } */
| | ~
| | |
| | (7) 'tmp.word_state' leaks here; was allocated at (1)
|