On Fri, Sep 28, 2001 at 03:33:27AM -0400, Christopher Faylor wrote: >On Fri, Sep 28, 2001 at 03:27:00AM -0400, Christopher Faylor wrote: >>On Fri, Sep 28, 2001 at 11:05:34AM +0400, egor duda wrote: >>>Hi! >>> >>>Friday, 28 September, 2001 Christopher Faylor [EMAIL PROTECTED] wrote: >>> >>>>> 320 12080627 [main] ssh 7436 peek_pipe: /dev/piper, saw EOF >>>>> 261 12080888 [main] ssh 7436 peek_pipe: saw eof on '/dev/piper' >>>>> 238 12081126 [main] ssh 7436 fhandler_pipe::ready_for_read: returning 1 >>> >>>CF> As usual, I can't duplicate this problem. I did see one oddity in some of the >new >>>CF> code that I added. I'll check in a patch for that shortly. >>> >>>Your last checkin fixed both cvs+ssh and testsuite's kill02 (it was >>>blocking on pipe read too) problems for me. Thanks! >> >>Phew. I am now noticing a SEGV in zsh when it tries to run perl, though. >> >>Could that be due to pthread_mutex problems in passwd.cc and grp.cc? > >To answer my own question: No. It seems like it is related to CYGWIN=ntsec. >If I have this set, then zsh/fork/exec dies when starting perl, apparently in >malloc. If I don't have CYGWIN=ntsec, then everything works ok. > >I don't know why yet.
It's not related to ntsec. Adding *any* environment variable seems to trigger it. I'm going back to debugging this but I didn't want Corinna to think that there was a problem with the ntsec code. cgf
