On Sat, 13 Nov 2010, Mark Riordan wrote:

My program now terminates with an access violation (segfault to you UNIX types) after a few writes. Access vios occur at various places:

Thanks for your patience and I'm sorry we go this back and forth between all sorts of weird bugs and errors, but I do believe we will end up with a better design and better code in the end.

[libssh2] 0.848116 Conn: Writing -4029 bytes on channel 0/0, stream #0

I think this is the key to the problems and I also think I know why this happens.

_libssh2_channel_write() can now (since my bug fix from a few hours ago) return more data than what is passed in to it (if a previous call got EAGAIN) and when it does that, it may confuse the caller greatly and might trigger a counter wrap into negative.

Tricky.

--

 / daniel.haxx.se
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Reply via email to