[Spice-devel] spice-gtk hangs

[Damien Churchill]

On 8 September 2011 15:22, Marc-André Lureau <mlureau at redhat.com> wrote:
> Hi
>
> ----- Original Message -----
>> Hi there,
>>
>> I've created a simple viewer application in Python, using the
>> spice-gtk library. It seems to be suffering from an issue where it
>> will occasionally just hang the display, and the only way to get it
>> back is to close the application and re-connect. Is this something
>> anyone else has experienced at all? It seems to happen after some time
>> (hours) of use.
>
> Nobody reported this problem before, afaik.
>
> Can you reproduce it with a different client, like spicy or virt-manager.
>

spicy doesn't seem to play nicely with the clipboard, at least on the
machines I've tried, so I haven't been able to test that out. Haven't
tried out virt-manager yet as it doesn't support multiple displays, as
fair as I know anyway.

> Are you building spice-gtk from source?
>

I'm building Ubuntu packages from source (0.7) since there aren't any
official ones that I can see, or at least up to date copies.

> Feel free to file a detailed bug report in bugzilla.freedesktop.org
>

I set spice_util_set_debug to True in my program and here is an
extract of the output leading up to when it hangs:

(vviewer:19988): GSpice-DEBUG: spice-channel-cache.h:91 cache_find:
image 59eab89e246f0529 [not found]
(vviewer:19988): GSpice-DEBUG: spice-channel-cache.h:107 cache_add:
image 59eab89e246f0529 (175)
(vviewer:19988): GSpice-DEBUG: channel-cursor.c:331 set_cursor: type
0, 1c, 41x39, flags 4, size 0
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x123, id 50343, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x123, id 50344, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x123, id 50345, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x123, id 50346, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x123, id 50347, ref 50257
(vviewer:19988): GSpice-DEBUG: channel-display.c:872
display_handle_stream_create: id 49
(vviewer:19988): GSpice-DEBUG: channel-display.c:925 scheduling next
stream render in 391 ms
(vviewer:19988): GSpice-DEBUG: channel-cursor.c:331 set_cursor: type
0, 1f, 41x39, flags 4, size 0
(vviewer:19988): GSpice-DEBUG: channel-cursor.c:331 set_cursor: type
0, 1f, 41x39, flags 4, size 0
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header:
980x197, id 50348, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header: 24x38,
id 50349, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header: 440x22,
id 50350, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header: 440x22,
id 50351, ref 50257
(vviewer:19988): GSpice-DEBUG: decode-glz.c:368 decode_header: 440x22,
id 50352, ref 50257

(vviewer:19988): GSpice-CRITICAL **: stream_mjpeg_data: assertion `j
== st->mjpeg_cinfo.rec_outbuf_height' failed
(vviewer:19988): GSpice-DEBUG: channel-display.c:925 scheduling next
stream render in 32 ms
(vviewer:19988): GSpice-DEBUG: channel-cursor.c:331 set_cursor: type
0, 1c, 41x39, flags 4, size 0
(vviewer:19988): GSpice-DEBUG: channel-cursor.c:331 set_cursor: type
0, 1c, 41x39, flags 4, size 0
Improper call to JPEG library in state 205


I note that there are some mjpeg fixes shortly after the 0.7 release
in the git commit log, would it be worth back-porting those to see how
that works?