Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: geom
-
Labels:None
-
Story Points:0.5
-
Epic Link:
-
Team:Data Release Production
Description
Some builds are failing on macOS in test_polynomials. It seems that the test is unstable in that it sometimes passes and sometimes fails.
Example failure:
Failed test output:
|
tests/test_polynomials
|
|
|
Running 9 test cases...
|
a=0.2272165537183282, b=0.2272165537183142, diff=1.398881011027697e-14 > 1.009044198056236e-14
|
tests/test_polynomials.cc:398: error: in "simplified2d": check compare(a1, b1, 100*DEFAULT_RTOL) has failed
|
a=-0.01667616780189118, b=-0.01667616780189041, diff=7.771561172376096e-16 > 7.405706182468739e-16
|
tests/test_polynomials.cc:398: error: in "simplified2d": check compare(a1, b1, 100*DEFAULT_RTOL) has failed
|
a=0.0221159430012603, b=0.02211594300126229, diff=1.998401444325282e-15 > 9.821451652519151e-16
|
tests/test_polynomials.cc:398: error: in "simplified2d": check compare(a1, b1, 100*DEFAULT_RTOL) has failed
|
a=0.0221159430012598, b=0.02211594300126181, diff=2.012279232133096e-15 > 4.910725826259466e-16
|
tests/test_polynomials.cc:402: error: in "simplified2d": check compare(a1, b1, 50*DEFAULT_RTOL) has failed
|
a=-0.02074950336545145, b=-0.02074950336544967, diff=1.77635683940025e-15 > 9.214630554344155e-16
|
tests/test_polynomials.cc:398: error: in "simplified2d": check compare(a1, b1, 100*DEFAULT_RTOL) has failed
|
a=-0.0207495033654517, b=-0.02074950336545071, diff=9.853229343548264e-16 > 4.60731527717222e-16
|
tests/test_polynomials.cc:402: error: in "simplified2d": check compare(a1, b1, 50*DEFAULT_RTOL) has failed
|
|
|
*** 6 failures are detected in the test module "polynomials"
|
Master does seem to build fine some of the time. Running the test gives a different answer each time so there seems to be a repeatability issue.
Attachments
Issue Links
- is triggered by
-
- Done
Activity
I can confirm this fixes my problem.
This is ready for review; anyone interested and available to do so should, as this is currently broken on master. I'd especially like a reviewer who could try building this on OS X, but if I can't get one soon I'll merge without review on the assumption that "probably fixed" is better than "definitely broken".