bpo-46765: Replace Locally Cached Strings with Statically Initialized Objects

[ericsnowcurrently]

https://bugs.python.org/issue46765

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @ericsnowcurrently for commit efe208f 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[ericsnowcurrently]

Alrighty then! asyncio on Mac and refleaks...

# https://buildbot.python.org/all/#/builders/474/builds/435
test_email leaked [6, 6, 6] memory blocks, sum=18
test_fractions leaked [2, 2, 2] references, sum=6
test_http_cookies leaked [2, 2, 2] references, sum=6
test_property leaked [6, 6, 6] references, sum=18
test_tracemalloc leaked [1, 1, 1] references, sum=3
test_zoneinfo leaked [12, 12, 12] references, sum=36
# https://buildbot.python.org/all/#/builders/18/builds/447
test_email leaked [7, 6, 6] memory blocks, sum=19
test_fractions leaked [2, 2, 2] references, sum=6
test_http_cookies leaked [2, 2, 2] references, sum=6
test_property leaked [6, 6, 6] references, sum=18
test_tracemalloc leaked [1, 1, 1] references, sum=3
test_zoneinfo leaked [12, 12, 12] references, sum=36
+
test_importlib leaked [67, 63, 63] references, sum=193
test_importlib leaked [36, 34, 34] memory blocks, sum=104
# https://buildbot.python.org/all/#/builders/411/builds/436
test_email leaked [6, 6, 6] memory blocks, sum=18
test_fractions leaked [2, 2, 2] references, sum=6
test_http_cookies leaked [2, 2, 2] references, sum=6
test_property leaked [6, 6, 6] references, sum=18
test_tracemalloc leaked [1, 1, 1] references, sum=3
test_zoneinfo leaked [12, 12, 12] references, sum=36
+
test_typing leaked [1, 1, 1] memory blocks, sum=3

[markshannon]

The asycnio test on Mac seems to be flaky. I've seen a failure on my PRs, which was fixed by re-running the tests.

The refleaks failures are puzzling. Are they passing on main?

[ericsnowcurrently]

The asycnio test on Mac seems to be flaky. I've seen a failure on my PRs, which was fixed by re-running the tests.

I suspected that but wasn't sure. It failed on repeated CI runs and on that buildbot. So I as at least going to investigate further.

The refleaks failures are puzzling. Are they passing on main?

Hmm, looks like they weren't. So maybe I'm okay here. I'll have to re-run it after the refleaks fix lands.

[erlend-aasland]

Hmm, looks like they weren't. So maybe I'm okay here. I'll have to re-run it after the refleaks fix lands.

FYI, they landed with GH-31389 just recently.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @ericsnowcurrently for commit efe208f 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[erlend-aasland]

I believe you will need to sync with main first:

$ git fetch --all
$ git merge --no-ff upstream/main
$ git push origin global-objects-use-pyid

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @ericsnowcurrently for commit 1852340 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[ericsnowcurrently]

That looks better. I'm still going to double-check that macOS failure before merging.

[sweeneyde]

I tested ./python -m test test_decimal -R3:3 and got [1, 1, 1] reference leaks on main, and bisected that to here. Is there some reason why that could happen on my machine but not on buildbots? I tested on windows with MSVC and also on WSL.

1f455361ecfb1892e375bdbee813cdf095b6cfb8 is the first bad commit
commit 1f455361ecfb1892e375bdbee813cdf095b6cfb8
Author: Eric Snow <ericsnowcurrently@gmail.com>
Date: Tue Feb 22 17:23:51 2022 -0700
[bpo-46765](https://bugs.python.org/issue46765): Replace Locally Cached Strings with Statically Initialized Objects (gh-31366)
https://bugs.python.org/issue46765

[ericsnowcurrently]

I'll take a look.

[sweeneyde]

Running git clean -f ended up fixing everything, so probably nothing to worry about.