I think the Github action for building CPython on Ubuntu is accidentally caching the built Python files.

If we take a look at: https://github.com/python/cpython/runs/455936632#step:7:1
and
https://github.com/python/cpython/pull/18567/checks?check_run_id=457662461#step:8:1

It seems like it's running way too fast (and producing too little output) to actually be building all of CPython.

Adding Steve who originally authored the action to the nosy list to see if they might have any insight.

46s seems okay to me, especially since the 23s for autoconf is separate.

Try introducing a warning/error into the source code and see if it breaks.

Aah sorry. This was my mistake, I didn't notice that we're passing `-s` (silent) to make. It is actually performing the build underneath, just not with the outputs that get produced form other CIs.

Steve, is there a reason the Github action and Azure pipelines:

* https://github.com/python/cpython/blob/1246d892038a693304549f8574e6c2784b91589a/.azure-pipelines/posix-steps.yml#L23

* https://github.com/python/cpython/blob/1246d892038a693304549f8574e6c2784b91589a/.github/workflows/build.yml#L58

both pass `-s` to make but Travis doesn't? It seems like it might potentially hide useful information about the build.

* https://github.com/python/cpython/blob/1246d892038a693304549f8574e6c2784b91589a/.travis.yml#L170

Probably just where I copied it from. PRs welcome!

New changeset 6aa1f1ecf7142a4117eedb8c570f30da1598616c by Ammar Askar in branch 'master':
bpo-39699: Don't silence make on Azure and Github CIs (GH-18583)
https://github.com/python/cpython/commit/6aa1f1ecf7142a4117eedb8c570f30da1598616c

New changeset 343bc06d8047e4a2e675394528dbb5155be1b3b5 by Steve Dower in branch '3.8':
bpo-39699: Don't silence make on Azure and Github CIs (GH-18583)
https://github.com/python/cpython/commit/343bc06d8047e4a2e675394528dbb5155be1b3b5

New changeset 3bf9de2fb954bd1131f4f41517d7d5c316fb95f8 by Steve Dower in branch '3.7':
bpo-39699: Don't silence make on Azure and Github CIs (GH-18583)
https://github.com/python/cpython/commit/3bf9de2fb954bd1131f4f41517d7d5c316fb95f8

Had to disable the doctests in the 3.7 branch because I just couldn't get them passing. Pablo did the work for 3.8 and later though, so they're on for those.

The text "TEST CHANGE TO BE UNDONE" is currently visible here: https://docs.python.org/3/whatsnew/3.0.html

It was introduced in commit 343bc06d8047e4a2e675394528dbb5155be1b3b5.

Should this test change have been undone?

Thanks for finding this Mark, looks like this was accidentally introduced in the 3.8 backport for testing the changes: https://github.com/python/cpython/pull/18670

The master (6aa1f1ecf7142a4117eedb8c570f30da1598616c) and 3.7 (3bf9de2fb954bd1131f4f41517d7d5c316fb95f8) commits look clean.

Opened up a quick pull request to fix it: https://github.com/python/cpython/pull/21089

New changeset 6eab52ffadb2836adb59d0578c84d247f05e19b1 by Ammar Askar in branch '3.8':
bpo-39699: Remove accidentally committed test change (GH-21089)
https://github.com/python/cpython/commit/6eab52ffadb2836adb59d0578c84d247f05e19b1