You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
I've noticed that when running coverage in parallel mode on containerized CI systems with heavy concurrency it happens sometimes that the generated suffix is not unique (depleted entropy pool would be my guess).
Describe the solution you'd like
A simple solution would be to allow the user to specify an optional additional suffix (either via cli flag or environment variable) since most CI systems already provide a unique job-id (or similar).
Describe alternatives you've considered
More elaborate schemes of automatically generating the suffix (e.g. taking a hash of the environment, incorporating date/time). But those seem either a bit too complicated or could also suffer from the same problem in high concurrency situations.
The text was updated successfully, but these errors were encountered:
Hmm, interesting point: coverage.py uses the process id to disambiguate, but docker uses the same pids over and over. My inclination would be to have a config value that is used as the entire suffix, but I think you can do everything you need now by setting data_file with an environment variable?
Is your feature request related to a problem? Please describe.
I've noticed that when running coverage in
parallel
mode on containerized CI systems with heavy concurrency it happens sometimes that the generated suffix is not unique (depleted entropy pool would be my guess).Describe the solution you'd like
A simple solution would be to allow the user to specify an optional additional suffix (either via cli flag or environment variable) since most CI systems already provide a unique job-id (or similar).
Describe alternatives you've considered
More elaborate schemes of automatically generating the suffix (e.g. taking a hash of the environment, incorporating date/time). But those seem either a bit too complicated or could also suffer from the same problem in high concurrency situations.
The text was updated successfully, but these errors were encountered: