On 03/11/2013 12:46 PM, Kornel Benko wrote:
Am Montag, 11. März 2013 um 15:33:36, schrieb Jean-Marc Lasgouttes
<lasgout...@lyx.org>
> Le 11/03/2013 14:42, Kornel Benko a écrit :
> > ==11211== LEAK SUMMARY:
> > ==11211== definitely lost: 33,700 bytes in 189 blocks
> > ==11211== indirectly lost: 59,454 bytes in 1,606 blocks
> > ==11211== possibly lost: 15,783 bytes in 59 blocks
> > ==11211== still reachable: 1,721,311 bytes in 8,392 blocks
> > ==11211== suppressed: 0 bytes in 0 blocks
> >
> > ==11211== Rerun with --leak-check=full to see details of leaked memory
>
> > I see, I should rerun, oh well ... this may take a while ...
This time it took 2:30 h to end, and another 2 min to shut down.
The output of
# valgrind --track-origins=yes --leak-check=full --log-file ... lyx
is somewhat lengthy.
> I am not sure there is so many leaks to find...
Yes, but probably the biggest would be enough. Nonetheless, there is
nothing bigger than 15kb.
Does not explain, what I was seeing if run alone. Rerunning again, no
such big leak. Maybe, I should
add InsetBibtex::plaintext() again.
Can you post a backtrace showing where that is being called from? I
don't know what advanced search is doing exactly, but I know it involves
creating a temporary file of some sort.
Yet another reason to want to disentangle the plaintext output routine
from anything used internally.
Richard
