On Thu, 2006-05-11 at 11:58 -0700, Mark Gilbert wrote:
> Because we finalize elsewhere. See this:
> http://www.abisource.com/bonsai/cvsview2.cgi?diff_mode=context&whitespace_mode=show&subdir=abi/src/af/xap/unix&command=DIFF_FRAMESET&file=xap_UnixApp.cpp&rev1=1.99&rev2=1.100&root=/cvsroot
Hmmm, imo, the code shouldn't be split up like this (over multiple
files). I would at least avoid the confusion I had.
> You see how if the version is at least 2.3, we use entirely both the
> new initialization routines and _the new finalization routines_? This
> cleaned up some issues that valgrind helped point out, as well as some
> issues with newer versions of FC and their cache management.
/me nods
> If your heap profiler tells you that the memory isn't taken care of,
> it's most likely because it's too limited in scope or is being given
> too constrictive parameters in the name of resource usage (the latter
> being more likely). Let me know if any further clarification is
> needed.
Valgrind complains about a lot of lost memory ... Will look into this
further.
Thanks for the explanation
Marc
Received on Fri May 12 14:54:24 2006
This archive was generated by hypermail 2.1.8 : Fri May 12 2006 - 14:54:24 CEST