Christophe Demory fixed the check to work better for non-blocking on HP-UX
systems. Bug report #701749.

removed the local variables for emacs and vim, use the new sample.emacs
way for emacs, and vim users should provide a similar non-polluting style

copyright year update in the source header

when sending an error message to the debugfunction, we append a newline so
that the output looks better

return -1 even if SSL_pending() doesn't return non-zero, as we don't really
care how many bytes that is readable NOW. Philippe Raoult reported the
bug in 7.10.3-pre3.

Fixed the usage of SSL_read() to properly return -1 if the EWOULDBLOCK
situation occurs, which it previously didn't!

This was reptoed by Evan Jordan in bug report #653022.

Also,

Fixed the usage of SSL_read() to properly return -1 if the EWOULDBLOCK
situation occurs, which it previously didn't!

This was reptoed by Evan Jordan in bug report #653022.

Also, if ERROR_SYSCALL is returned from SSL_write(), include the errno number
in the error string for easier error detection.

show more ...

Lehel Bernadt's fix to prevent debug message to get sent on errors when
debug wasn't enabled

failf() now sends the text to the debug function callback

updated source code boilerplate/header

Markus Oberhumer improved an out-of-memory check
I reformatted some functions using a different indent than the rest
of the file.

made VERBOSE output more like it used to be, HEADER_IN is thus also ignored
by the internal debugfunction callback

Jean-Philippe Barrette-LaPierre provided his patch that introduces
CURLOPT_DEBUGFUNCTION and CURLOPT_DEBUGDATA.

error code cleanup, use the new SEND/RECV errors

Clarence Gardner pointed out the not-taken-care-of return codes from SSL_read

copyright string (year) update

sendf() now deals with Curl_write() returning -1 properly, which it might
do if the write would've blocked

Kevin Roth's discovered SSL download problem

Georg Horn the previous SSL_read() fix, this was actually the fix I did
on my test machine! :-)

Georg Horn found yet another SSL reading problem caused by the non-blocks.
This was a real bummer!

Added #include <errno.h>

Andr�s Garc�a patched. It now checks for EWOULDBLOCK properly on windows
boxes.

SSL writes passed back a silly length...

Curl_read() and Curl_write() are both now adjusted to return properly in
cases where EWOULDBLOCK or equivalent is returned. We must not block.

Curl_read() now returns a negative return code if EWOULDBLOCK or similar

this seems to correct the SSL reading problem introduced when switching
over to non-blocking sockets, but this loops very nastily. We should return
back to the select() and wait there until m

this seems to correct the SSL reading problem introduced when switching
over to non-blocking sockets, but this loops very nastily. We should return
back to the select() and wait there until more data arrives, not just blindly
attempt again and again...

show more ...