On Wed, 5 Aug 1998, Aaron Nabil wrote: > Mike Wade writes... > >Aug 5 16:35:20 news1 nntpcache-client[28770]: sockets.c:446: <- ARTICLE > >1740260Aug 5 16:35:20 news1 nntpcache-client[28770]: sockets.c:482: => > >[library.airnews.net.] ARTICLE 1740260 > >Aug 5 16:35:20 news1 nntpcache-client[28817]: sockets.c:345 failed > >assertion: scfg > >Aug 5 16:35:20 news1 nntpcache-client[28817]: nntpcache.c:281: SIGSEGV! > > > >Any ideas? > > It's possible the server went away. I wrote a patch that might help with Probably not. That error means that Cfget() was called with a NULL pointer for the scfg structure. That isn't supposed to be possible. If the server had gone away, scfg->fd would be invalid. In this case, scfg doesn't even exist. It could maybe result from an insane configuration in nntpcache.servers. How often does this problem occur? What operating system? What compiler (as it could be an compiler optimzing bug)? > this, article.c does a write of the "ARTICLE 1740260" blindly and doesn't > notice if the server died until much later. I've had this cause some Yes, but is is handled later. > really cool error messages. (how does "fatal: /dev/zero: mmap failed: Resource > temporarily unavailable" grab ya?) Really? I'm not sure how an invalid socket fd would cause a mmap() failure. I would likely take that error at face-value: you hit some per-process or per-user limit. Tom
