Bug 108924 - memory leak in doloop_warn
Summary: memory leak in doloop_warn
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: fortran (show other bugs)
Version: 13.0
: P3 normal
Target Milestone: ---
Assignee: anlauf
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-02-24 12:31 UTC by Richard Biener
Modified: 2023-03-03 14:04 UTC (History)
1 user (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2023-02-24 00:00:00

Attachments
Tentative patch (256 bytes, patch)
2023-02-24 16:37 UTC, anlauf 		Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Richard Biener 2023-02-24 12:31:42 UTC 
valgrind leak check complains

==8747== 16 bytes in 2 blocks are definitely lost in loss record 6 of 2,309
==8747==    at 0x4C346A4: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==8747==    by 0x5697698: __gmp_default_allocate (in /usr/lib64/libgmp.so.10.3.2)
==8747==    by 0x56A7A22: __gmpz_init_set (in /usr/lib64/libgmp.so.10.3.2)
==8747==    by 0x874110: do_subscript (frontend-passes.cc:2863)
==8747==    by 0x874110: do_function(gfc_expr**, int*, void*) (frontend-passes.cc:2710)
==8747==    by 0x870675: gfc_expr_walker(gfc_expr**, int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5243)
==8747==    by 0x870859: gfc_expr_walker(gfc_expr**, int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5250)
==8747==    by 0x870859: gfc_expr_walker(gfc_expr**, int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5250)
==8747==    by 0x8727D9: gfc_code_walker(gfc_code**, int (*)(gfc_code**, int*, void*), int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5680)
==8747==    by 0x872857: gfc_code_walker(gfc_code**, int (*)(gfc_code**, int*, void*), int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5687)
==8747==    by 0x872857: gfc_code_walker(gfc_code**, int (*)(gfc_code**, int*, void*), int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5687)
==8747==    by 0x872857: gfc_code_walker(gfc_code**, int (*)(gfc_code**, int*, void*), int (*)(gfc_expr**, int*, void*), void*) (frontend-passes.cc:5687)
==8747==    by 0x8738CB: doloop_warn(gfc_namespace*) (frontend-passes.cc:3079)

so it looks like some gmp vars are not released.
Comment 1 anlauf 2023-02-24 16:37:29 UTC 
Created attachment 54532 [details]
Tentative patch

Does the attached fix this?  Without a testcase this seems to be the fix
obtained from reviewing the surrounding code.  Regtests cleanly here.
Comment 2 GCC Commits 2023-02-24 20:30:58 UTC 
The master branch has been updated by Harald Anlauf <anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:45f406c4f62e516b58dcda20b5a7aa43ff0aa0f3

commit r13-6336-g45f406c4f62e516b58dcda20b5a7aa43ff0aa0f3
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Fri Feb 24 19:56:32 2023 +0100

    Fortran: frontend passes do_subscript leaks gmp memory [PR108924]
    
    gcc/fortran/ChangeLog:
    
            PR fortran/108924
            * frontend-passes.cc (do_subscript): Clear used gmp variable.
Comment 3 anlauf 2023-02-24 20:39:59 UTC 
Taken.  Waiting on feedback.
Comment 4 Richard Biener 2023-03-03 14:04:26 UTC 
Building the benchmark set fprate_any_fortran of SPEC CPU 2017 there's no leak
from an allocation in do_subscript or do_function anymore.