On Mon, 30 Mar 2009, Simon Josefsson wrote:
>> A) valgrind slows down the execution a lot. I can get the leak to occur in my
>> tests but it seems virtually impossible to make happen when valgrind
>> monitors/slows down the code. A plain memory-leak detection would be
>> almost no extra overhead.
>
> I would worry about A) indicating some more worry-some problem though, it
> smells like a race issue?
Yes that's a risk, or I'm also thinking it could be a case where the
non-blocking code and its states aren't done right so that we need an EAGAIN
to happen at the "right" point for the problem to trigger and that of course
requires some pretty specific timing.
-- / daniel.haxx.se ------------------------------------------------------------------------------ _______________________________________________ libssh2-devel mailing list libssh2-devel_at_lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libssh2-develReceived on 2009-03-30