Fix tests (hopefully)

[dlfivefifty]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.03%. Comparing base (f2fe965) to head (42b877f).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #182      +/-   ##
==========================================
+ Coverage   82.32%   83.03%   +0.71%     
==========================================
  Files          11       12       +1     
  Lines        1420     1509      +89     
==========================================
+ Hits         1169     1253      +84     
- Misses        251      256       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dlfivefifty]

@DanielVandH I think this has hit a bug in the UL code?

[dlfivefifty]

oh wait.... the failure is actually in infqr

[dlfivefifty]

Hmm... there's a bug in banded matrix broadcasting that is only triggered on MacOS on CI that's breaking a lot of tests.... not sure how to debug it.

[DanielVandH]

Weird. I don't believe my code touched any of that. The only relevant recent change would be JuliaArrays/LazyArrays.jl@052a0f6 I think? But nothing there looks like it would ever cause some OS-dependent issue. The manifests between each OS are the same too so that's weird.

[dlfivefifty]

Right it couldn't have been your code and this bug has been popping up elsewhere. I've tried setting random seeds, etc., but nothing makes it reliably triggered.

All the banded broadcast should be Julia code. So it shouldn't be an issue with BLAS or LAPACK.

The fact that its an out-of-band error kind of rules at the possibility of an array not being set values. Why would this impact sparsity?

So I'm at a complete loss. I just hope it's some weird issue with the CI setup but I can't imagine what it would be. I've successfully debugged CI bugs by adding print statements but I don't think I have the motivation right now to investigate it...

[DanielVandH]

I've had some array access issues like this before with DelaunayTriangulation.jl with CI bugs that are completely non-existent locally. It took a lot of print debugging and care to get the environment correctly setup to somehow identify the cause (and now the tests just include thousands of random tests everywhere which is a bit annoying)... if it continues to be a problem I can try and look into it if necessary.