Project

General

Profile

Bug #2828

some osmo-trx tests failing in i586 OBS machine

Added by pespin 9 months ago. Updated 2 months ago.

Status:
Feedback
Priority:
Normal
Assignee:
Category:
-
Target version:
-
Start date:
01/13/2018
Due date:
% Done:

50%

Spec Reference:

Description

Tests run fine in x86_64 but not in x86.

## ---------------------------------- ##
## osmo-trx 0.2.0.28-8639 test suite. ##
## ---------------------------------- ##

Regression tests.

  1: BitVectorTest                                   ok
  2: InterthreadTest                                 ok
  3: LogTest                                         ok
  4: PRBSTest                                        ok
  5: SocketsTest                                     ok
  6: TimevalTest                                     FAILED (testsuite.at:38)
  7: VectorTest                                      ok
  8: convolve_test                                   FAILED (testsuite.at:50)

https://build.opensuse.org/package/live_build_log/network:osmocom:nightly/osmo-trx/Debian_9.0/i586

I can reproduce the 2 test failures in my i686 netbook. I attach the output of the testsuite.dir in here.

testsuite.tar.gz testsuite.tar.gz 12 KB pespin, 01/14/2018 09:46 PM

Related issues

Related to OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4New2018-01-11

History

#1 Updated by pespin 9 months ago

#2 Updated by pespin 9 months ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 50

Timeval test should be fixed in already merged osmo-trx 28ce315a3225188211c2ed52844716f88093035e.

#3 Updated by pespin 9 months ago

For the convolve_test, there seems to be small rounding issues between code used in here (i686) and x86_64:

-float y[] = {0.389293,10.824917,-0.676577,10.619646,0.283489,11.279525,0.384482,11.586230,0.711259,11.540458,-0.391531,11.281723,0.019900,12.278080,-0.070459,11.104558,0.087938,11.825965,-1.003252,11.698885,0.358887,11.911197,-0.678904,11.933812,0.245140,11.886644}
+float y[] = {0.389292,10.824915,-0.676577,10.619645,0.283489,11.279525,0.384482,11.586229,0.711259,11.540459,-0.391531,11.281723,0.019899,12.278080,-0.070459,11.104559,0.087938,11.825966,-1.003252,11.698885,0.358887,11.911197,-0.678904,11.933812,0.245140,11.886644}

-float y[] = {5.354852,5.387001,4.829278,5.046340,5.849788,5.775999,5.653334,5.372714,5.999860,5.593828,5.628739,5.178002,6.010774,6.186034,6.337766,5.538046,5.616131,6.289612,5.486091,5.835261,6.277413,5.894117,5.563587,6.082063,5.828556,6.160175}
+float y[] = {5.354852,5.387001,4.829278,5.046340,5.849788,5.775999,5.653334,5.372714,5.999860,5.593828,5.628739,5.178002,6.010775,6.186034,6.337765,5.538045,5.616131,6.289611,5.486091,5.835262,6.277413,5.894117,5.563587,6.082062,5.828556,6.160175}

#4 Updated by pespin 9 months ago

  • Related to Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4 added

#5 Updated by pespin 9 months ago

  • Priority changed from Urgent to Normal

As the Timeval test was fixed and the convolve_test one is disabled for non x86_64 buildhosts, I decrease priority to normal since it can build fine.

I also sent an e-mail to the mailing list and ttsou stated he was going to look at #2826 in a short time.
https://lists.osmocom.org/pipermail/openbsc/2018-January/011655.html

#6 Updated by pespin 2 months ago

  • Status changed from In Progress to Feedback
  • Assignee changed from pespin to ttsou

Hi ttsou, any news regarding this topic and #2826?

Also available in: Atom PDF

Add picture from clipboard (Maximum size: 48.8 MB)