Closed paraynaud closed 1 year ago
Base: 99.07% // Head: 99.07% // No change to project coverage :thumbsup:
Coverage data is based on head (
11e3338
) compared to base (303275a
). Patch has no changes to coverable lines.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
@dpo, I had a failed test on some of my last PRs. I used to re-run the failed test, but i wanted to fix it before v0.2.0.
The error came from the partitioned linear-operator based on Vector (and not on PartitionedVector). This operator is made to check that solutions from CG for PartitionedVectors and Vectors are the same. Generally it works fine, but sometimes, for an unknown reason, the following code:
will return
x_vector
with every component asNaN
.I made additional tests and
return some
NaN
whileMatrix(epm)
doesn't. More generallypm_v * rand(n)
may return someNaN
(i don't know why). It is really strange because if you runpm_v * rand(n)
twice, you may not have the same result. One may haveNan
s while the other doesn't.To avoid failure of continuous integration tests, this PR check that
x_vector
(result of CG for Vectors) doesn't haveNaN
before comparing it with the result of CG for PartitionedVectors.