Hardhat coverage behaves different locally vs in GitHub ci

[LyonSsS]

Context: In order to validate coverage of tests we implemented a coverage mechanism which does execute coverage locally via hardhat and then in GitHub CI and if the CI job is != local file to fail as the file was tampered.
Still we had multiple jobs that executed with different results from local to Github CI jobs.

Here are few examples: https://github.com/sprintertech/sprinter-stash-contracts/actions/runs/20311697276/job/58344394052
Just fork from this commit then try run it locally and compare with the CI results. (https://github.com/sprintertech/sprinter-stash-contracts/actions/runs/20311697276/job/58344394052#step:9:924 )

[cgewecke]

Hi, thanks for opening.

Can you show me a comparison screenshot of:

• the terminal table printed on your local machine at the end of the coverage run
• the terminal table printed in CI for the same code state

Are these different?

[LyonSsS]

left is local right is ci

[cgewecke]

It seems like something is actually different about execution in the two contexts, i.e the coverage utility is probably reporting a real difference between the environments.

If I was going to debug this (in the worst case) I'd try to:

• make the html report artifact produced by coverage in CI visible somehow. One way is to upload it from CI to a free temporary hosting service like surge.sh
• compare that report with my local version and see what specific lines/tests are executing differently.
• reason about what might cause the difference based on your understanding of the tests:
  • are my local block starts different?
  • is my system clock is different?

You know the code well, maybe something jumps out at you just from the variance you see in the superficial report?