On 13/08/2014, at 11:50 AM, john skaller wrote: > > On 13/08/2014, at 4:09 AM, john skaller wrote: > >> Ok, I just managed to get Ctrl-C to terminate the webserver >> (after one more request, since the signal isn't detected >> whilst waiting to accept the next connection). >> >> After mucking around, several different runs with FLX_FINALISE=1 >> to force a terminating GC I get this: >> >> ==12449== LEAK SUMMARY: >> ==12449== definitely lost: 830 bytes in 24 blocks >> ==12449== indirectly lost: 0 bytes in 0 blocks >> ==12449== possibly lost: 0 bytes in 0 blocks >> ==12449== still reachable: 30,701 bytes in 22 blocks >> ==12449== suppressed: 0 bytes in 0 blocks >> >> To repeat, this is the same several runs in sequence. >> So there is no leak according to Valgrind. >> If there were a leak the 24 or 22 number would be different >> in different runs. But it's not. > > The logical conclusion: the (accumulating) leak is in the epoll demuxer > since there's no leak on my Mac (using kqueue) and that's > the only difference.
On Linux .. weird .. =13617== LEAK SUMMARY: ==13617== definitely lost: 830 bytes in 24 blocks ==13617== indirectly lost: 0 bytes in 0 blocks ==13617== possibly lost: 576 bytes in 2 blocks ==13617== still reachable: 264 bytes in 4 blocks ==13617== suppressed: 0 bytes in 0 blocks -- john skaller skal...@users.sourceforge.net http://felix-lang.org ------------------------------------------------------------------------------ _______________________________________________ Felix-language mailing list Felix-language@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/felix-language
