Question about automatic release detection

[rafikdraoui]

The python SDK uses git rev-parse --short HEAD to determine a release from a git repository. sentry-cli releases propose-version returns the full-length hash for HEAD.

If I were to use sentry-cli's proposed version to make a release when deploying, and let sentry-python automatically detect the release at runtime, would they be both referring to the same release? On the one hand, the hash are not exactly equal, so they could be considered different, but on the other hand, they semantically are "the same" according to git.

One way to find out would be to actually try it, but I still haven't configured releases on my project and thought I might try to see if that would be a viable option before going too deep into it 😃

[untitaker]

@mitsuhiko can you look into this? You added this functionality and know most about releases at the moment.

also cc @matejminar

[petarmaric]

Just ran into this same issue for my client - when using https://github.com/marketplace/actions/sentry-release within our GitHub actions-powered CD workflow to automatically create a new Sentry release we get an automatically detected release such as 365020454471 in the Sentry dashboard.

However, the current git-based automatic detection returns a different release string 3650204, for the same deployment.

Because of this we keep getting two "different" releases in the Sentry dashboard for the same deployment, where one receives errors/events (git-based automatic detection, on the server itself), but the other has relevant codebase/author/commits info (GitHub actions automatic detection, during deployment). Any idea how to go about this @untitaker and @mitsuhiko ?

[petarmaric]

ping @untitaker