Hi Daniel, Daniel Stenberg wrote: > On Sun, 9 Jul 2006, Harald Dunkel wrote: > >> My testcase did not crash anymore, but of course the problem is not >> _that_ much reproducable. > > Indeed, but you instead added a memory-leak. And I truly don't think > this work-around is good enough to actually hide the problems for you > properly... >
How many entries would you usually expect for the splay tree? I doubt that the memory leak is a serious problem, compared to the SEGV we have at the moment. >> Since the splay functions seem to be used internally only, this is >> surely just a workaround, if it helps at all. > > A better work-around, if you want work-arounds, is to simply make > Curl_expire() return immediately without doing anything at all. This too > will make curl_multi_timeout() break, but there aren't that many actual > users of that function (yet). > > Neither of these work-arounds actually fix anything though, they just > hide the problem. > Creating a copy of node is more defensive. Of course it is just hiding the problem. Using curl I wouldn't be interested in its internal details, anyway. Regards Harri
signature.asc
Description: OpenPGP digital signature
