MEMORY LEAKS

vicious · 2005-01-11 07:38:01

There have been a lot of leaks recently.

firefox, inkscape, mplayer.

Isn't this because of gtk 2.6?

sarah31 · 2005-01-11 12:02:12

could be if you know a way of pin pointing it it i suggest filing a bug. I would definitley agree that the report of memeory leaks is way up lately.

phrakture · 2005-01-11 15:34:52

sarah31 wrote:

could be if you know a way of pin pointing it it i suggest filing a bug. I would definitley agree that the report of memeory leaks is way up lately.

try running lint on some of the source... or maybe some memory profiling tools... I don't actually know of any beyond ccmalloc... anyone have good experience with any memory profilers?

edit:
valgrind looks very useful....
try running:

valgrind --tool=memcheck --leak-check=yes <program>

vicious · 2005-01-11 19:47:32

At least with inkscape memory profiles doesn't make any sense, because it uses a garbage collector.

However, I'm very curious why drawing an ellipse with inkscape leaks about 0,5MB of memory? That is very much!!!

IceRAM · 2005-01-11 20:11:10

I've tried running valgrind on inkscape and it doesn't work (valgrind --tool=memcheck --leak-check=yes inkscape).

==6389== Process terminating with default action of signal 11 (SIGSEGV)
==6389== Bad permissions for mapped region at address 0x52C00158
==6389== at 0x1C535800: GC_mark_from (in /usr/lib/libgc.so.1.0.2)
==6389== by 0x1C536FF4: GC_mark_some (in /usr/lib/libgc.so.1.0.2)
==6389== by 0x1C52E09A: GC_stopped_mark (in /usr/lib/libgc.so.1.0.2)
==6389== by 0x1C52EA53: GC_try_to_collect_inner (in /usr/lib/libgc.so.1.0.2)

Same error running as root. valgrind might need some special parameters.

iBertus · 2005-01-12 00:15:12

Almost everyone thats upgraded to gtk 2.6 has talked about having memory leak problems. If the leak is in a library used by a program will the profiler work?

Could it not be gtk 2.6 but another library that is common to all these apps?

IceRAM · 2005-01-12 00:24:48

iBertus wrote:

If the leak is in a library used by a program will the profiler work?

I guess so. Check my screenshots in this thread (profiling pacman):
http://bbs.archlinux.org/viewtopic.php?t=9166
libc calls are also listed.

Actually.. those screenshots were made using the callgrind tool (in valgrind).. so.. you can say a profiler works. Leaks showed in that topic belong to pacman and they were spotted because valgring somehow hangs to "malloc"s to check when memory is allocated and runs the code inside somehow of virtual machine to see when/if it is freed (this might not be the best technical description, but I know that valgrind works).

I'm mostly using KDE apps + Firefox (rarely gqview, ethereal)... and I can't say I experience these problems.

dtw · 2005-01-12 02:50:10

wtf IS a memory leak?

iBertus · 2005-01-12 04:34:09

dibblethewrecker wrote:

wtf IS a memory leak?

This is what programmers and other cool people call the sloppy allocation of memory by programs. When a program is "leaking" memory, it allocates memory but then fails to release it when finished. The result, which is very noticible with repeating processes, is a block of memory that becomes useless until it's reclaimed.

Back in Windows, I once wrote a buggy piece of GDI code that leaked a few KB every time the use would refresh the graph (it was a graphic calculator app). Ahh, memories.

phrakture · 2005-01-12 15:43:51

dibblethewrecker wrote:

wtf IS a memory leak?

iBertus has it right... if you understand some basic C, here's an example:

void *p = malloc(1024); //allocate 1Kb of memory
p = NULL; //hooray, I don't have a p pointer anymore
          // so I can't call free(p) to release the memory

vicious · 2005-01-12 16:06:33

The interesting thing is that inkscape uses a garbage collector. So it should reclaim all the leaked memory sooner or later. The leak in gtk 2.6 must be more subtle (and undectable automatically). It must store a pointer to the leaked memory somewhere, but not use the pointer anymore and forget to delete it.

phrakture · 2005-01-12 16:14:25

vicious wrote:

The interesting thing is that inkscape uses a garbage collector. So it should reclaim all the leaked memory sooner or later. The leak in gtk 2.6 must be more subtle (and undectable automatically). It must store a pointer to the leaked memory somewhere, but not use the pointer anymore and forget to delete it.

good call, yeah if you're using some GC (which I never trust) you have to maintain a reference to that pointer in order to have it leak...

dtw · 2005-01-13 02:49:25

cool - thanks guys

Arch Linux

#1 2005-01-11 07:38:01

MEMORY LEAKS

#2 2005-01-11 12:02:12

Re: MEMORY LEAKS

#3 2005-01-11 15:34:52

Re: MEMORY LEAKS

#4 2005-01-11 19:47:32

Re: MEMORY LEAKS

#5 2005-01-11 20:11:10

Re: MEMORY LEAKS

#6 2005-01-12 00:15:12

Re: MEMORY LEAKS

#7 2005-01-12 00:24:48

Re: MEMORY LEAKS

#8 2005-01-12 02:50:10

Re: MEMORY LEAKS

#9 2005-01-12 04:34:09

Re: MEMORY LEAKS

#10 2005-01-12 15:43:51

Re: MEMORY LEAKS

#11 2005-01-12 16:06:33

Re: MEMORY LEAKS

#12 2005-01-12 16:14:25

Re: MEMORY LEAKS

#13 2005-01-13 02:49:25

Re: MEMORY LEAKS

Board footer