Repository navigation
httpx CI group also runs httpx2 test suite #7151
Description
Looks like the httpx target in the Networks group in CI is also testing httpx2.
Open httpx here, for instance: https://github.com/getsentry/sentry-python/actions/runs/31385230658/job/93444058506 It only contains httpx2 runs. Both httpx and httpx2 are run in the httpx group (I assume because we do/tox does some prefix matching). In Python 3.14, specifically, there are no httpx versions to run the tests against, so we only run them against httpx2.
This is not a big deal, but we could check whether we can isolate the test suites.
Activity
linear-code commentedon Aug 10, 2026
PY-2727
[-]httpx CI group runs httpx2 test suite[/-][+]httpx CI group also runs httpx2 test suite[/+]ZaafirRizwan commentedon Aug 15, 2026
Hi! I'd like to work on this.
I traced the collision to
scripts/runtox.sh: the generated Network workflow invokespy<version>-httpxandpy<version>-httpx2separately, butruntox.shuses an unanchored substring match overtox -l. As a result,py3.14-httpxselects the existingpy3.14-httpx2-*environments even though there are no Python 3.14httpxenvironments.I found the same prefix-collision precedent in #4730, where
openaiwas renamed toopenai_baserather than changing the intentionally broad selector.Would you prefer the same approach here—rename the tox suite factor to something like
httpx1orhttpx_base, retainintegration_name: "httpx", update the Network group, and regeneratetox.iniand the workflows—or would you rather makeruntox.shboundary-aware?I'll wait for confirmation on the preferred approach before implementing it.
sentrivana commentedon Aug 17, 2026
Hey @ZaafirRizwan. Generally, if there's a precedent I'd prefer we follow that, so in this case we could go with renaming the
httpxgroup to something else, maybehttpx0since it's still in 0.x?It'll need the additional
integration_name, and maybe more -- check what we did in the analogous openai case.Feel free to go ahead and give this a shot.
Epsilon006 commentedon Sep 5, 2026
Hi @ZaafirRizwan — are you still working on this? I don’t see an open PR linked yet. If you’ve stepped away, I’d be happy to take it.
I’ll follow @sentrivana’s suggested precedent: rename the legacy httpx tox group (likely to httpx0), add integration_name, and mirror the analogous openai-base setup. I’ll reproduce the generated CI matrix locally first and keep the change focused.
ZaafirRizwan commentedon Sep 5, 2026
Hey @Epsilon006, thanks for checking in! Still on it. PR #7398 is up, taking exactly that approach (
httpx0+integration_name, mirroringopenai-base).ci: Isolate `httpx` test suite from `httpx2` (#7398)