Codecov Report

Attention: Patch coverage is 72.72727% with 6 lines in your changes missing coverage. Please review.

Project coverage is 74.10%. Comparing base (929c3d6) to head (dbe2fa1).
Report is 2 commits behind head on main.

@@           Coverage Diff           @@
##             main    #1904   +/-   ##
=======================================
  Coverage   74.09%   74.10%           
=======================================
  Files         444      444           
  Lines       55516    55514    -2     
  Branches     9128     9128           
=======================================
+ Hits        41137    41138    +1     
+ Misses      11287    11285    -2     
+ Partials     3092     3091    -1     

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

Some background here: In the .NET assembly, I want to use a module initializer to set the log callback. A module initializer ensures some initialization is run before any user code. In order to achieve this initialization previously, I was using a static constructor in the the Application class. This has two drawbacks:

1. All user access to Cantera through the .NET interface has to be routed through the Application class to ensure the static constructor is run.
2. The runtime has to emit runtime checks on each access of the Application class to ensure the static constructor has run. (This can be seen by the lack of the `Not so for simple static field initializers, which are prevented from referencing other fields. See https://learn.microsoft.com/en-us/dotnet/fundamentals/code-analysis/quality-rules/ca1810)

The second one will likely only cause a minimal case in every-day use, but the first requires a very awkward library design.

When switching to a module initializer, I found that logging tests started failing. It turns out this was because the runtime was running the initializer on a different thread from the one the tests were running on. Thus, the logCallback was not getting invoked because it was only applied to the thread the on which the initializer run. By applying this commits on this PR, the test started working again.

Fwiw, the PR appears to have broken the post-merge Cython test, although it’s never clear whether it’ll wash out as the pre-release progresses
https://github.com/Cantera/cantera/actions/runs/15519808401/job/43691507167

From my reading of that error log, I don't think it's anything in this PR that broke that test, but a change in the pre-release Cython. It is a bit concerning though, if there are changes in Cython that are going to break our somewhat non-standard practice of compiling multiple Cython source files and linking them as a single extension module.