[dmoroian]

Hi it seems that both import utilities fluentMeshToFoam/fluent3DMeshToFoam have a memory leak.
The error appears in both release 1.4.1 and development 1.4.1-dev distribution.
Here is a small sample msh file that produces this error: fluentMeshToFoamError
Here is the valgrind log for fluentMeshToFoam:

==18676== Memcheck, a memory error detector.
==18676== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==18676== Using LibVEX rev 1804, a library for dynamic binary translation.
==18676== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==18676== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==18676== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==18676== For more details, rerun with: -v
==18676==
==18676== My PID = 18676, parent PID = 8547. Prog and args are:
==18676== fluentMeshToFoam
==18676== ./
==18676== disc
==18676== geom/disc.msh
==18676==
==18676== Invalid free() / delete / delete[]
==18676== at 0x4A1F96E: free (vg_replace_malloc.c:323)
==18676== by 0x614DD1A: free_mem (in /lib64/libc-2.4.so)
==18676== by 0x614D991: __libc_freeres (in /lib64/libc-2.4.so)
==18676== by 0x491C31C: _vgnU_freeres (vg_preloaded.c:60)
==18676== by 0x608E2F4: exit (in /lib64/libc-2.4.so)
==18676== by 0x607915A: (below main) (in /lib64/libc-2.4.so)
==18676== Address 0x4048630 is not stack'd, malloc'd or (recently) free'd
==18676==
==18676== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 9 from 2)
==18676== malloc/free: in use at exit: 8 bytes in 1 blocks.
==18676== malloc/free: 595,120 allocs, 595,120 frees, 27,887,659 bytes allocated.
==18676== For counts of detected errors, rerun with: -v
==18676== searching for pointers to 1 not-freed blocks.
==18676== checked 630,312 bytes.
==18676==
==18676== LEAK SUMMARY:
==18676== definitely lost: 8 bytes in 1 blocks.
==18676== possibly lost: 0 bytes in 0 blocks.
==18676== still reachable: 0 bytes in 0 blocks.
==18676== suppressed: 0 bytes in 0 blocks.
==18676== Rerun with --leak-check=full to see details of leaked memory.

and similarly for fluent3DMeshToFoam:

==18677== Memcheck, a memory error detector.
==18677== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==18677== Using LibVEX rev 1804, a library for dynamic binary translation.
==18677== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==18677== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==18677== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==18677== For more details, rerun with: -v
==18677==
==18677== My PID = 18677, parent PID = 8547. Prog and args are:
==18677== fluent3DMeshToFoam
==18677== ./
==18677== disc
==18677== geom/disc.msh
==18677==
==18677== Invalid free() / delete / delete[]
==18677== at 0x4A1F96E: free (vg_replace_malloc.c:323)
==18677== by 0x614DD1A: free_mem (in /lib64/libc-2.4.so)
==18677== by 0x614D991: __libc_freeres (in /lib64/libc-2.4.so)
==18677== by 0x491C31C: _vgnU_freeres (vg_preloaded.c:60)
==18677== by 0x608E2F4: exit (in /lib64/libc-2.4.so)
==18677== by 0x607915A: (below main) (in /lib64/libc-2.4.so)
==18677== Address 0x4048630 is not stack'd, malloc'd or (recently) free'd
==18677==
==18677== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 9 from 2)
==18677== malloc/free: in use at exit: 8 bytes in 1 blocks.
==18677== malloc/free: 173,856 allocs, 173,856 frees, 9,157,314 bytes allocated.
==18677== For counts of detected errors, rerun with: -v
==18677== searching for pointers to 1 not-freed blocks.
==18677== checked 630,312 bytes.
==18677==
==18677== LEAK SUMMARY:
==18677== definitely lost: 8 bytes in 1 blocks.
==18677== possibly lost: 0 bytes in 0 blocks.
==18677== still reachable: 0 bytes in 0 blocks.
==18677== suppressed: 0 bytes in 0 blocks.
==18677== Rerun with --leak-check=full to see details of leaked memory.

Dragos

[mattijs]

Just ran it on the tutorials icoFoam/elbow/elbow.msh file in 1.4.1 and no problem. Something special about your case? Can you attach it or send it to me?

valgrind fluentMeshToFoam $FOAM_RUN/icoFoam elbow $FOAM_TUTORIALS/icoFoam/elbow/elbow.msh

..
End
==4387==
==4387== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 1)
==4387== malloc/free: in use at exit: 0 bytes in 0 blocks.
==4387== malloc/free: 0 allocs, 0 frees, 0 bytes allocated.
==4387== For counts of detected errors, rerun with: -v
==4387== All heap blocks were freed -- no leaks are possible.

[dmoroian]

Hi Mattijs,
As I put it above, the sample msh file can be downloaded from: fluentMeshToFoamError, and represents just 2 cylinders one on top of another. The first is discretized using hexahedra, whereas the second is discretized using tetrahedra:

Thank you for looking at it,
Dragos

[dmoroian]

Hmm,
I've just tried it on the elbow mesh, and I get the same error:

==22414== Memcheck, a memory error detector.
==22414== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==22414== Using LibVEX rev 1804, a library for dynamic binary translation.
==22414== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==22414== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==22414== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==22414== For more details, rerun with: -v
==22414==
==22414== My PID = 22414, parent PID = 21418. Prog and args are:
==22414== fluentMeshToFoam
==22414== ./
==22414== elbow
==22414== elbow/elbow.msh
==22414==
==22414== Invalid free() / delete / delete[]
==22414== at 0x4A1F96E: free (vg_replace_malloc.c:323)
==22414== by 0x614DD1A: free_mem (in /lib64/libc-2.4.so)
==22414== by 0x614D991: __libc_freeres (in /lib64/libc-2.4.so)
==22414== by 0x491C31C: _vgnU_freeres (vg_preloaded.c:60)
==22414== by 0x608E2F4: exit (in /lib64/libc-2.4.so)
==22414== by 0x607915A: (below main) (in /lib64/libc-2.4.so)
==22414== Address 0x4048630 is not stack'd, malloc'd or (recently) free'd
==22414==
==22414== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 9 from 2)
==22414== malloc/free: in use at exit: 8 bytes in 1 blocks.
==22414== malloc/free: 111,616 allocs, 111,616 frees, 6,025,695 bytes allocated.
==22414== For counts of detected errors, rerun with: -v
==22414== searching for pointers to 1 not-freed blocks.
==22414== checked 630,344 bytes.
==22414==
==22414== LEAK SUMMARY:
==22414== definitely lost: 8 bytes in 1 blocks.
==22414== possibly lost: 0 bytes in 0 blocks.
==22414== still reachable: 0 bytes in 0 blocks.
==22414== suppressed: 0 bytes in 0 blocks.
==22414== Rerun with --leak-check=full to see details of leaked memory.

I'm using the latest available version of valgrind at this date (3.3.0 Current Release)
What version of valgrind are you using?

Dragos

[mattijs]

valgrind 3.2.3 and 3.3.0 both give 0 errors. Maybe there is a difference in the (f)lex versions and the bug is in the flex part. Feel free to debug it ;-)

[dmoroian]

Ok, thanks for the support, I'll try.
Though the flex part should not play a role here since the realease version 1.4.1 is not compiled by me (I've donwloaded the binaries together with the sources).

Dragos

[dmoroian]

Hello again,
I did some more tests on different machines, both at the office and at home (all 64bit). On all of them, I get the same result from valgrind:

==22414== LEAK SUMMARY:
==22414== definitely lost: 8 bytes in 1 blocks.

I've asked a friend to do a similar test. He has 1.4 release version, and when he tried it he got also a memory leak error:

==23077== LEAK SUMMARY:
==23077== definitely lost: 241,295 bytes in 8,326 blocks.

To summarize:
- OpenFoam-1.4.1 on Suse 10.1 64 bit
- OpenFoam-1.4.1-dev on Suse 10.1 64 bit
- OpenFoam-1.4.1 on OpenSuse 10.3 64 bit
- OpenFoam-1.4.1-dev on OpenSuse 10.3 64 bit
- OpenFoam-1.4 on CentOS 4.6 (32 bit OS on 64 bit proc)
They all show a memory leak in valgrind:
valgrind fluentMeshToFoam ./ elbow elbow/elbow.msh
Anyone out there can please confirm this? Just run the above command.

Thanks,
Dragos

[dmoroian]

A new test on Linux evs-008 2.6.18-5-686 #1 SMP Wed Oct 3 00:12:50 UTC 2007 i686 GNU/Linux
Debian GNU/Linux 4.0 with the same command:
valgrind fluentMeshToFoam ./ elbow elbow/elbow.msh
gives the following:

==3604== LEAK SUMMARY:
==3604== definitely lost: 160 bytes in 12 blocks.

Anyone else?

[adam_mitchell]

I to get this problem on Fedora Core release 6 (Zod)

valgrind fluentMeshToFoam ./ elbow elbow/elbow.msh

==22808== Invalid free() / delete / delete[]
==22808== at 0x4A0548E: free (vg_replace_malloc.c:233)
==22808== by 0x3594702DAA: free_mem (in /lib64/libc-2.5.so)
==22808== by 0x35947029C1: __libc_freeres (in /lib64/libc-2.5.so)
==22808== by 0x4802314: _vgnU_freeres (vg_preloaded.c:60)
==22808== by 0x3594632EA4: exit (in /lib64/libc-2.5.so)
==22808== by 0x359461DA4A: (below main) (in /lib64/libc-2.5.so)
==22808== Address 0x773EAC0 is not stack'd, malloc'd or (recently) free'd
==22808==
==22808== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 5 from 1)
==22808== malloc/free: in use at exit: 8 bytes in 1 blocks.
==22808== malloc/free: 111,669 allocs, 111,669 frees, 5,932,062 bytes allocated.
==22808== For counts of detected errors, rerun with: -v
==22808== searching for pointers to 1 not-freed blocks.
==22808== checked 599,960 bytes.
==22808==
==22808== LEAK SUMMARY:
==22808== definitely lost: 8 bytes in 1 blocks.
==22808== possibly lost: 0 bytes in 0 blocks.
==22808== still reachable: 0 bytes in 0 blocks.
==22808== suppressed: 0 bytes in 0 blocks.
==22808== Use --leak-check=full to see details of leaked memory.

[mattijs]

Both 1.4.1 and 1.5beta give me:

==27496== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 7 from 1)

with --leak-check=full it reports that the two leaks are in
- yyFlexLexer::yyensure_buffer_stack() and
- getpwuid_r@@GLIBC_2.2.5
neither I think are anything to do with us.

Running 4.2.1 & 4.2.2 gcc, OpenSuse 10.2, 64bit mode, double precision.

[dmoroian]

It seems that you're right Mattijs, I found on the internet several posts related to this error (but on different topics: Memory leak in flex generated file). There are also some work around's, so as soon as I have time I will try one of them.
...but first I have to learn flex and its capability.

Dragos