Commit 683aa0f6b52fb1014873c961709102b5006372fc made send_existing() send
more than just the second part of a packet when the kernel did not accept
the full packet, but the function still overlooked the SSH protocol
overhead in each packet, often 48 bytes.
If only the last few bytes of a packet remained, then the packet would
erroneously be considered completely sent, and the next call to write
more data in the session would return a -39 error.
When libssh2_transport_write() is called to continue sending a
partially sent packet the write direction flag must not be cleared
until the previous packet has been completely sent, or the app would
hang if the packet still isn't sent completely, since select() gets
called by the internal blocking emulation layer in libssh2 but would
then not be watching the socket for writability.
Clear the flag only once processing of previous packet data is
complete and a new packet is about to be prepared.
The channel layer sends packets using the transport layer, possibly
calling _libssh2_transport_write() many times for each packet.
The transport layer uses the send_existing() helper to send out any
remaining parts of previous packets before a new packet is started.
The bug made send_existing() consider the entire packet sent as soon as it
successfully sent the second part of a packet, even if the packet was not
completely done yet.
I made this change just to easier grep for "return .*EAGAIN" cases
as they should be very rare or done wrongly. Already worked to find
a flaw, marked with "TODO FIXME THIS IS WRONG" in channel.c. I also
fixed a few cases to become more general returns now when we have
more unified return codes internally.
Alexander Lamaison tracked down that my previous commit broke SFTP
reads in some aspects. The reversion now gets back to always recv()
until EAGAIN is returned so that the code no longer treats a short
read as an indication that it is "enough for now".
The bad commit in particular had another independent change included,
which is to clear the direction-bits first in the transport read
and write functions, but this reversion does not revert that change.
Clearing those bits first is a good thing.
The anonymous bug report #2822910 pointed out that debugdump() was
stupidly called within the send_existing() function. At closer
inspection was the report not only right, but it also revealed
another problem to me: when the _libssh2_send() function returns
after sending only a part of the buffer, it would then misbehave.
This probably is very rare though, which must be the reason we
haven't seen a bigger problem with this.
as we're already doing the correct check further down anyway there's no point
in doing the (wrong) check further up as well. Paul Veldkamp pointed this out.
and introduced a transport.h header.
* Fixed the blocking mode to only change behavior not the actual underlying
socket mode so we now always work with non-blocking sockets. This also
introduces a new rule of thumb in libssh2 code: we don't call the
external function calls internally. We use the internal (non-blocking)
ones!
* libssh2_channel_receive_window_adjust2 was added and
libssh2_channel_receive_window_adjust is now deprecated
* Introduced "local" header files with prototypes etc for different parts
instead of cramming everything into libssh2_priv.h. channel.h is the
first.
properly after recv() and send() calls (that internally are now known as
_libssh2_recv() and _libssh2_send()) so that the API and more works fine on
windows too!
some tests:
* cut off "_ex" from several internal function names
* corrected some log outputs
* simplified libssh2_channel_read_ex() and made it much faster in the process
* cut out {{{ and }}} comments that were incorrect anyway
* fixed sftp_packet_ask() to return the correct packet by using memcmp() and
not strncmp()
* fixed mkdir()'s wait for packet to use the correct request_id - it
semi-worked previously because strncmp() in sftp_packet_ask() made it
match far too easily.
* took away the polling functionality from sftp_packet_ask() since it wasn't
used
libssh2_session_block_directions() which returns a bitmask for what
directions the connection blocks. It is to be used applications that use
non-blocking sockets and when a libssh2 function returns
LIBSSH2_ERROR_EAGAIN this function can be used to figure out in which
direction the socket would block and thus it can wait for the socket to
again be ready for communication in that direction before it calls libssh2
again.
all the way up to the user interface. All code modules bug sftp.c have
been completed.
Functions that return an "int", or similar return LIBSSH2CHANNEL_EAGAIN to
indicate some part of the call would block, in non-blocking mode.
Functions that return a structure, like "LIBSSH2_CHANNEL *", return NULL
and set the libssh2 error. The error can be obtained with either
libssh2_session_last_error() or libssh2_session_last_errno(). Either of
these will return the error code of LIBSSH2_ERROR_EAGAIN if the
call would block, in non-blocking mode.
The current state of a function and some variable are keep in the
structures so that on the next call the operation that would block can
be retried again with the same data.
return an error. "errno" should have been zero by the recv() call but
doesn't seem to be on all systems. Only check when recv() says there
is an error.
Bug: 1703467
Patch: 1703468
the recent commits converted the tabs to 4 spaces, which matched the
initial indent size. Other commits converted the tabs to 8 spaces, this
didn't match.
All the code has been converted to 4 space indents. No changes to line
lengths or actual code was performed. This is in preperation to my up
coming non-blocking work so my commits should only be code changes and
line lengths in the code I am working on.
various stuff if libssh2 was built with debug. If built without debug, the
function does nothing.
2 - configure --enable-debug is now enough to build a debug version (including
picky compiler options)
3 - internally, we no longer need/use #ifdef/#endif around all uses of the
_libssh2_debug() function/macro.
The scp.c example is the first application to test this new debug logging.
blocking way. The channel code is now responsible for enabling/disabling
blocking status and to work with it.
I've also modified indenting and fixed compiler warnings at places, and
added a bunch of new examples in example/simple that I've used to verify that
the code still runs like before.
libssh2_channel_{read|write}nb_ex() and libssh2_sftp_{read|write}nb() are the
four new functions that supposedly work non-blocking.
