error with jpeg encoding

[Timo Juhani Lindfors]

Unhandled error while processing packet from peer
Traceback (most recent call last):
  File "python/xpra/protocol.py", line 238, in _process_packet
    self._process_packet_cb(self, decoded)
  File "python/xpra/client.py", line 1023, in process_packet
    self._packet_handlers[packet_type](self, packet)
  File "python/xpra/client.py", line 908, in _process_draw
    window.draw(x, y, width, height, coding, data)
  File "python/xpra/client.py", line 214, in draw
    self._backing.draw_rgb_image(gc, x, y, width, height, gtk.gdk.RGB_DITHER_NONE, data)
IndexError: rgb_buf is not large enough

when trying --encoding jpeg --jpeg-quality 60

[Antoine Martin]

Hmm, that's the mmap drawing codepath, which does not use jpeg at all: when mmap is detected the data is just transfered as-is.

Is this repeatable? What do I need to do?
Did it recover afterwards? Or did you have to restart the client?

According to ​this post, this may be because the "rowstride" is different from "w*3" - no idea why though, as mmap doesn't go anywhere near w or rowstride.
What resolution and bit depth are you running on the client?
Could also be that the backing got resized to be smaller than the data that got sent, which would be easier to deal with. Does that sound plausible?

[Antoine Martin]

if the feature is supported, send the rowstride with the pixel data

[Antoine Martin]

The patch above should probably be merged no matter what: although I don't think the rowstride!=w*3 codepath is hit very often at the moment, it might in the future (at which point the rowstride re-alignment should get re-written in cython for speed.
I am now getting the error, must be a recent svn change.

[Antoine Martin]

Lots of fixes, including this bug in r389

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/65