Re: [BUG] unit/test-midi spurious failure

Hi Felipe,

Am 28.05.2017 um 19:55 schrieb Felipe Tonello:

Hi Stefan,

On Fri, May 26, 2017 at 9:17 AM, Stefan Seyfried
[off-list ref] wrote:

quoted

Hi all,

I get spurious failures in unit/test-midi

quoted

ERROR:unit/test-midi.c:287:compare_events: assertion failed (ev1->data.ext.len == ev2->data.ext.len): (5 == 4)

quoted

ERROR:unit/test-midi.c:287:compare_events: assertion failed (ev1->data.ext.len == ev2->data.ext.len): (1024 == 1023)
Aborted (core dumped)
ERROR:unit/test-midi.c:287:compare_events: assertion failed (ev1->data.ext.len == ev2->data.ext.len): (5 == 4)
Aborted (core dumped)

About one in four runs fails:

#!/bin/bash
G=0; B=0;
for i in `seq 1 1000`; do
  if unit/test-midi; then
    G=$((G+1)
  else
    B=$((B+1))
  fi
done
echo "good: $G bad: $B"

good: 730 bad: 270

I have no idea what this means, if this is an error in the
test code or a sign something is wrong with the testing system.

It started to happen with bluez-5.45, still happens with git master.

Does it happen on BlueZ 5.44? If so, then I believe it is a bug in the
MIDI code.

I have not seen it in BlueZ 5.44, only in 5.45 and master.
But because it is a spurious failure, I might have just been lucky (I
only notice failing builds in the openSUSE Build Service if I actively
look for them, but if some dependencies are rebuilt, then the bluez
package also gets automatically rebuilt and the spurious failure might
"fix" itself)

Thanks for finding it out.

"make check" is part of the openSUSE RPM packaging process, to find such
things, so I did not have to do any real work to find this :-)

Best regards,

	Stefan
-- 
Stefan Seyfried

"For a successful technology, reality must take precedence over
 public relations, for nature cannot be fooled." -- Richard Feynman