test-display-streaming creates mismatching output
Hi,
in addition to the build time make check Debian/Ubuntu used to run a later test using ./test-display-streaming
as documented in README.
The upgrade from 14.0
to 14.2
required some adaptions to how the tests were run which now is like this test that is executed in the unpacked source.
But I happened to find it reporting me:
TESTER: listening on port 5912 (unsecure)
attache_worker
set_compression_level
set_client_capabilities: present 0 caps 0
req_cursor_notification
req_cursor_notification
[...]
wrote screen shot to output.ppm
FAIL: Regressions were found!
Please, take a look in your code and go fix it!
(process:19988): Spice-WARNING **: 16:45:29.470: Connection reset by peer
That made me wonder why it worked before, but it turned out that the old code always failed to call ./regression-test.py
(the retval of execve was not checked in the past) not being part of the distributed tarball anymore.
So the old tests never actually compared the output on Debian/Ubuntu but by me fixing up the availability of ./regression-test.py
and base_test.ppm
the test now actually ran that comparison - and now reported a fail.
base_test.ppm is from git master server/tests/base_test.ppm while output.ppm is what my tests seem to reliably create - a mistmatch as detected by the tests.
I now wonder: a) is this test not meant to be run on the release tarballs (as the files are missing there) b) if the overall test is ok, why would our execution create broken images (but spice otherwise seems to work fine) c) should we just remove this follow on test and stick to the build time checks already used
I'm asking for your guidance to fix (or drop) our test accordingly.