On Aug 20 18:23, Haojun Bao wrote: > On Thu, Aug 20, 2009 at 4:39 PM, Corinna > Vinschen<corinna-cyg...@cygwin.com> wrote: > > On Aug 20 14:09, Haojun Bao wrote: > >> I have done some debugging, and the culprit should be dup(2) syscall. > >> Here's another test case, this time written in C. > >> > >> Note that the cygheap_start and cygheap_max value will be very likely > >> different on your computer, so you should use gdb to take a peek into > >> cygwin1.dll to get your value. Or else you should remove the reference > >> to these memory location. > >> > >> The test case will show cygheap is always growing, and at the end it will > >> print > >> 1 [main] a 3560 Q:\a.exe: *** fatal error - cmalloc would have returned > >> NULL > > > > Thanks for the testcase! It was pretty easy to find the culprit with > > it. Deep in dup(), the strings for the filenames of the new file > > descriptor were allocated twice. While I was at it, I also found two > > other potential memory leaks, which would just show up much less > > frequent. > > > > This fixes your find testcase as well, and probably (hopefully?) also > > the problem reported in http://cygwin.com/ml/cygwin/2009-08/msg00620.html > > > > I applied a patch to CVS and will upload a -60 release asap. > > > > > > Thanks again, > > Corinna > > > Great. In fact, I also found this code myself might cause problem in path.h: > (we should test if path is NULL, and free it before the memcpy, and > other member pointers should also be checked and free-ed first, is it > about right?:-)
Yes and no. It might look cleaner to free the pointers at this point if they are non-NULL, but in fact the operator= is called after the path_conv content has been memcpy'ed to another fhandler. So, if you free the pointers, you free the pointers of another file descriptor. SEGV's galore! Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader cygwin AT cygwin DOT com Red Hat -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
