autopkgtests flaky on s390x
Bug #1832360 reported by
Olivier Tilloy
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
libreoffice (Ubuntu) |
Fix Released
|
Medium
|
Marcus Tomlinson |
Bug Description
The "uno" autopkgtest has been failing quite a few times recently, and this seems to have gotten worse on eoan (it happened on disco already). See http://
### float does not match! failed
struct comparison test failed
### float does not match! failed
recursive test results failed
standard test failed
exception occurred: error: test failed!
I'm attaching the relevant excerpt of the most recent failed test run.
Changed in libreoffice (Ubuntu): | |
status: | New → In Progress |
importance: | Undecided → Medium |
no longer affects: | df-libreoffice |
To post a comment you must log in.
So turns out the test case is being a little naughty by comparing floats with a "==", which s390x in particular does not handling so well.
You can see in the debug log that the test float value is set to 17.0815f but actually gets stored as '17.081499', '17.081501' and '17.081503' in various places.
Ideally the test should be comparing floats by difference within the bounds of an epsilon.
Preparing a patch for this now.