Skip to content

Skip optimisations to speedup test_freeze rebuild - #157264

Draft
hugovk wants to merge 1 commit into
python:mainfrom
hugovk:3.16-skip-test_freeze-opt
Draft

Skip optimisations to speedup test_freeze rebuild#157264
hugovk wants to merge 1 commit into
python:mainfrom
hugovk:3.16-skip-test_freeze-opt

Conversation

@hugovk

@hugovk hugovk commented Sep 10, 2026

Copy link
Copy Markdown
Member

For LTO buildbots, test_tools from test_freeze.py accounts for the majority of the test runtime:

Builder (build) Test step test_tools %
PPC64LE CentOS9 LTO 35m09s 31m09s 89%
PPC64LE Fedora Stable LTO 23m53s 17m51s 75%
PPC64LE Fedora Rawhide LTO 21m26s 20m57s 98%
AMD64 CentOS9 LTO 14m02s 13m59s 100%
AMD64 Fedora Rawhide LTO 13m24s 11m18s 84%
aarch64 CentOS9 LTO 11m19s 10m54s 96%
AMD64 Fedora Stable LTO 10m01s 7m14s 72%
s390x RHEL9 LTO 8m46s 8m06s 92%
aarch64 Fedora Rawhide LTO 8m26s 7m10s 85%
s390x Fedora Stable LTO 8m08s 7m37s 94%
aarch64 CentOS10 LTO 7m59s 6m26s 81%
s390x Fedora Rawhide LTO 7m44s 6m01s 78%
aarch64 Fedora Stable LTO 7m39s 6m47s 89%
s390x CentOS10 LTO 7m10s 6m22s 89%

Often the rest of the test suite had finished running, and we're just hanging on for this one to finish.

This is because it involves rebuilding the interpreter, and this includes the LTO flags, which results in a slow rebuild. The rebuild is to test the freezing machinery, and optimisation flags aren't needed here.

See also #103053 which disabled the skipped the test on PGO, but not LTO. We could skip, but let's see if this improves things enough.

@hugovk

hugovk commented Sep 10, 2026

Copy link
Copy Markdown
Member Author

!buildbot LTO PR

@bedevere-bot

Copy link
Copy Markdown

🤖 New build scheduled with the buildbot fleet by @hugovk for commit fcecce3 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F157264%2Fmerge

The command will test the builders whose names match following regular expression: LTO PR

The builders matched are:

  • AMD64 Fedora Stable LTO PR
  • aarch64 CentOS10 LTO PR
  • s390x Fedora Rawhide LTO PR
  • PPC64LE Fedora Stable LTO PR
  • s390x Fedora Stable LTO PR
  • AMD64 Fedora Rawhide LTO PR
  • aarch64 Fedora Rawhide LTO PR
  • s390x CentOS10 LTO PR
  • s390x RHEL9 LTO PR
  • PPC64LE CentOS9 LTO PR
  • PPC64LE Fedora Rawhide LTO PR
  • aarch64 Fedora Stable LTO PR
  • aarch64 CentOS9 LTO PR
  • AMD64 CentOS9 LTO PR

@hugovk

hugovk commented Sep 10, 2026

Copy link
Copy Markdown
Member Author

Results look good:

Builder Before After
Test step test_tools % Test step test_tools %
PPC64LE CentOS9 LTO 35m09s 31m09s 89% 10m26s 8m04s 77%
PPC64LE Fedora Stable LTO 23m53s 17m51s 75% 10m22s 10m06s 97%
PPC64LE Fedora Rawhide LTO 21m26s 20m57s 98% 11m56s 8m32s 72%
AMD64 CentOS9 LTO 14m02s 13m59s 100% 7m26s 5m58s 80%
AMD64 Fedora Rawhide LTO 13m24s 11m18s 84% 7m37s 4m14s 56%
aarch64 CentOS9 LTO 11m19s 10m54s 96% 3m13s 3m03s 95%
AMD64 Fedora Stable LTO 10m01s 7m14s 72% 4m04s 3m37s 89%
s390x RHEL9 LTO 8m46s 8m06s 92% pending
aarch64 Fedora Rawhide LTO 8m26s 7m10s 85% 3m27s 3m00s 87%
s390x Fedora Stable LTO 8m08s 7m37s 94% 6m34s 2m54s 44%
aarch64 CentOS10 LTO 7m59s 6m26s 81% 3m06s 2m48s 90%
s390x Fedora Rawhide LTO 7m44s 6m01s 78% pending
aarch64 Fedora Stable LTO 7m39s 6m47s 89% 3m43s 2m37s 70%
s390x CentOS10 LTO 7m10s 6m22s 89% pending

test_tools still accounts for much of the test step's time, but it's much quicker now.

The s390x buildbots are offline until the 15th due to migration from a Z15 to a Z17 machine, so will need to compare before and after on the new machine when it's ready.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants