“Interesting” mtests results on some platforms

• Feb 7, 2019 - 21:29
Reported version
3.0
Type
Functional
Frequency
Once
Severity
S3 - Major
Reproducibility
Always
Status
active
Regression
No
Workaround
No
Project

For example: https://buildd.debian.org/status/fetch.php?pkg=musescore-snapshot&arch=…

--- /<>/mtest/capella/io/test5.cap-ref.mscx 2019-01-30 12:34:27.000000000 +0000
+++ test5.cap.mscx 2019-02-07 02:10:06.180188142 +0000
@@ -8,8 +8,8 @@
       8
       12
       1
-      0.666667
-      1.8
+      -0.00616776
+      -194.56

     1
     1

… is one of the differences shown. Possible causes:

  • char may be signed or unsigned by default
  • differences in floating point arithmetics
  • compiler bugs (of course, won’t avoid them)
  • code bugs ;) (see above for two likely causes)
  • testcase bugs ;) (see above for likely causes)

Not a regression though: 2.3.2 behaves the same: https://buildd.debian.org/status/fetch.php?pkg=musescore&arch=powerpc&v… (see https://buildd.debian.org/status/package.php?p=musescore for build logs, click on the “Installed” in the Status column to access individual build logs).

This likely means that MuseScore will fail (crash or, worse, corrupt data) on some platforms.


Comments

See https://buildd.debian.org/status/package.php?p=musescore-snapshot&suite… for build logs of the last snapshots (currently tracking releases) in Debian; in the same vain, click on “Installed” in the “Status” line to access the build log (the “all” and one of them, usually “amd64”, will not be available because these were the architectures I uploaded, but if there is interest, I can mail build logs someplace when I upload, before I discard them).