Ticket #713 (closed defect: fixed)

Opened 9 years ago

Last modified 9 years ago

mapraster memory leaks

Reported by: mturk@… Owned by: warmerdam
Priority: high Milestone: 4.4 release
Component: MapServer C Library Version: 4.3
Severity: normal Keywords:
Cc:

Description 

Mapraster has a memory leak drawing tiled raster layers caused by not properly
freeing temporary layer.
The memory leak was tested using Softpixel MemCheckDeluxe v1.2.2
It can be found at http://prj.softpixel.com/mcd/, and it quite usefull and easy
to  include in any project (also it has BSD style license).

Attachments

mapraster.patch.txt Download (5.0 KB) - added by mturk@… 9 years ago.
Patch for mapraster.c resolving memory leaks

Change History

Changed 9 years ago by mturk@…

Patch for mapraster.c resolving memory leaks

Changed 9 years ago by dmorissette

  • cc steve.lime@…, morissette@… added
  • owner changed from sdlime to fwarmerdam 
Reassigned to Frank, the owner of mapraster.c

Changed 9 years ago by fwarmerdam

  • status changed from new to assigned
  • milestone set to 4.4 release 
This needs to be reviewed and applied in time for 4.4.  The tricky part is
reproducing the leak.  From the patch it would appear the leaking only 
occurs in error handling cases.

Changed 9 years ago by fwarmerdam 

I did a run with one of the raster tile files running under valgrind and it
was clean, no leaks.  But I ran with the tile index itself missing and got 
the following:

==2222== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 64 from 1)
==2222== malloc/free: in use at exit: 91858 bytes in 11 blocks.
==2222== malloc/free: 1376 allocs, 1365 frees, 743710 bytes allocated.
==2222== For counts of detected errors, rerun with: -v
==2222== searching for pointers to 11 not-freed blocks.
==2222== checked 19440216 bytes.
==2222== 
==2222== 5 bytes in 1 blocks are still reachable in loss record 1 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x4DD8FF: strdup (in /lib/tls/libc-2.3.3.so)
==2222==    by 0x80C89D5: msDrawRasterLayerLow (mapraster.c:1303)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 9 bytes in 1 blocks are still reachable in loss record 2 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x4DD8FF: strdup (in /lib/tls/libc-2.3.3.so)
==2222==    by 0x8055086: initLayer (mapfile.c:2156)
==2222==    by 0x80C89BD: msDrawRasterLayerLow (mapraster.c:1300)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 16 bytes in 1 blocks are still reachable in loss record 3 of 11
==2222==    at 0x1B9043FD: calloc (vg_replace_malloc.c:176)
==2222==    by 0x5BB308: _dlerror_run (in /lib/libdl-2.3.3.so)
==2222==    by 0x5BAED0: dlsym (in /lib/libdl-2.3.3.so)
==2222==    by 0x1C358C6F: __errno_location (vg_libpthread.c:2127)
==2222==    by 0x1B996296: pj_init (pj_init.c:233)
==2222==    by 0x8051552: msProcessProjection (mapfile.c:912)
==2222==    by 0x805BCD6: initMap (mapfile.c:4005)
==2222==    by 0x805C75C: loadMapInternal (mapfile.c:4189)
==2222==    by 0x805D7CD: msLoadMap (mapfile.c:4407)
==2222==    by 0x804DB75: main (shp2img.c:80)
==2222== 
==2222== 
==2222== 20 bytes in 1 blocks are still reachable in loss record 4 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x4DD8FF: strdup (in /lib/tls/libc-2.3.3.so)
==2222==    by 0x80C89FF: msDrawRasterLayerLow (mapraster.c:1305)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 80 bytes in 1 blocks are still reachable in loss record 5 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x8050F8D: msInitProjection (mapfile.c:745)
==2222==    by 0x8054F91: initLayer (mapfile.c:2142)
==2222==    by 0x80C89BD: msDrawRasterLayerLow (mapraster.c:1300)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 164 bytes in 1 blocks are still reachable in loss record 6 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x80AB007: initHashTable (maphash.c:71)
==2222==    by 0x80551A8: initLayer (mapfile.c:2185)
==2222==    by 0x80C89BD: msDrawRasterLayerLow (mapraster.c:1300)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 280 bytes in 1 blocks are still reachable in loss record 8 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x8055216: initLayer (mapfile.c:2197)
==2222==    by 0x80C89BD: msDrawRasterLayerLow (mapraster.c:1300)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 524 bytes in 1 blocks are definitely lost in loss record 9 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x80C8970: msDrawRasterLayerLow (mapraster.c:1295)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== 
==2222== 2560 bytes in 1 blocks are still reachable in loss record 10 of 11
==2222==    at 0x1B903BF7: operator new(unsigned) (vg_replace_malloc.c:133)
==2222==    by 0x756FB5A: std::__default_alloc_template<true,
0>::_S_chunk_alloc(unsigned, int&) (in /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x756FA6C: std::__default_alloc_template<true,
0>::_S_refill(unsigned) (in /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x756F767: std::__default_alloc_template<true,
0>::allocate(unsigned) (in /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x75750F7: std::string::_Rep::_S_create(unsigned,
std::allocator<char> const&) (in /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x7575E4E: (within /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x75725A8: std::string::string(char const*, std::allocator<char>
const&) (in /usr/lib/libstdc++.so.5.0.5)
==2222==    by 0x1BDEDF01: __static_initialization_and_destruction_0(int, int)
(/usr/include/c++/3.3.3/bits/stl_alloc.h:652)
==2222==    by 0x1BDEDFC9: _GLOBAL__I__ZN3DDS9duplicateERKS_ (in
/debian/home/warmerda/gdal/libgdal.so)
==2222==    by 0x1BED5108: (within /debian/home/warmerda/gdal/libgdal.so)
==2222== 
==2222== 
==2222== 88000 bytes in 1 blocks are still reachable in loss record 11 of 11
==2222==    at 0x1B903A80: malloc (vg_replace_malloc.c:131)
==2222==    by 0x8054E41: initLayer (mapfile.c:2106)
==2222==    by 0x80C89BD: msDrawRasterLayerLow (mapraster.c:1300)
==2222==    by 0x8077ACB: msDrawRasterLayer (mapdraw.c:1116)
==2222==    by 0x8076402: msDrawLayer (mapdraw.c:754)
==2222==    by 0x8075548: msDrawMap (mapdraw.c:419)
==2222==    by 0x804E280: main (shp2img.c:200)
==2222== 
==2222== LEAK SUMMARY:
==2222==    definitely lost: 524 bytes in 1 blocks.
==2222==    possibly lost:   0 bytes in 0 blocks.
==2222==    still reachable: 91134 bytes in 9 blocks.
==2222==         suppressed: 200 bytes in 1 blocks.

So, in at least some error cases there are leaks.  I will try and patch
as suggested or as appropriate to correct the leaks - though to be honest
fixing leaks that occur only in error cases isn't super high priority.

Changed 9 years ago by fwarmerdam

  • status changed from assigned to closed
  • resolution set to fixed 
I have applied an approximate form of the supplied patch. 
I did not try and catch leak in some rare cases (ie. the old style raster
drivers).  All the leaks listed about were cleanuped in the case where the
whole tileindex is missing.
Note: See TracTickets for help on using tickets.