Bug 400783

Summary: valgrind complains about a leak in its own code
Product: [Developer tools] valgrind Reporter: Andrew Pennebaker <andrew.pennebaker>
Component: memcheckAssignee: Julian Seward <jseward>
Status: RESOLVED NOT A BUG    
Severity: normal CC: tom
Priority: NOR    
Version First Reported In: 3.14 SVN   
Target Milestone: ---   
Platform: Other   
OS: Linux   
Latest Commit: Version Fixed/Implemented In:
Sentry Crash Report:

Description Andrew Pennebaker 2018-11-07 05:17:15 UTC 
==1117== Invalid free() / delete / delete[] / realloc()
==1117==    at 0x4CB3F9B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==1117==    by 0x40717F3: reclaim_gaps (dynlink.c:508)
==1117==    by 0x4072D1E: load_library (dynlink.c:1061)
==1117==    by 0x4073893: load_preload (dynlink.c:1156)
==1117==    by 0x4073893: __dls3 (dynlink.c:1619)
==1117==    by 0x40734ED: __dls2 (dynlink.c:1441)
==1117==    by 0x4075BBF: ??? (in /usr/lib/libc.so)
==1117==    by 0x1: ???
==1117==    by 0x1FFF000DBE: ???
==1117==    by 0x1FFF000DC8: ???
==1117==  Address 0x4ebe180 is in a rw- mapped file /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so segment
Comment 1 Tom Hughes 2018-11-07 07:00:00 UTC 
What makes you think that is in it's own code, that look like it is in the C library or the dynamic linker?
Comment 2 Andrew Pennebaker 2018-11-07 16:09:15 UTC 
When I comment out all of my free()'s, valgrind continues to complain "Invalid free()", and also indicates sections in the valgrind code, not my code.
Comment 3 Tom Hughes 2018-11-07 16:30:14 UTC 
The only reference to valgrind in that stack trace is to our replacement free routine which has detected the problem - the actual problem is in the dynlink.c code about that which has apparently passed an invalid pointer to free.
Comment 4 Andrew Pennebaker 2018-11-07 19:33:19 UTC 
I do not see this behavior with earlier versions of valgrind such as 3.12. It’s worth pointing out that this fallacious free() warning is happening on Void Linux, compared to no fallacious warnings in Debian.
Comment 5 Tom Hughes 2018-11-08 00:13:20 UTC 
Look there is basically nothing we can do with the information you have provided  - the stack trace is incomplete because the symbols appear to have been stripped and you're using an obscure distribution that I doubt any of us have.

There is nothing in the information you have provided that indicates a bug in valgrind.

That is pretty much all I can say on the information currently available - maybe there is a bug in valgrind, maybe there isn't, but we have no way of determining that for sure and all indication are that there isn't.
Comment 6 Julian Seward 2019-03-10 08:15:01 UTC 
This is nothing to do with Valgrind.  Closing as invalid.  If you have
further information indicating that it is a bug in Valgrind, feel free
to reopen.