Uh oh!
There was an error while loading.Please reload this page.
- Notifications
You must be signed in to change notification settings - Fork34.1k
Comments
GH-124567: Reduce overhead of debug build for GC. Should help CI performance#126777
GH-124567: Reduce overhead of debug build for GC. Should help CI performance#126777markshannon merged 3 commits intopython:mainfrom
Conversation
markshannon commentedNov 13, 2024
Reduces the time it takes CI to run the |
pablogsal commentedNov 13, 2024
Not opposing to this but this will mean that debug builds that are redistributed won't have GC list validation so linked list corruption will be more difficult to debug when users leverage debug builds of the interpreter. |
markshannon commentedNov 13, 2024
This only removes the extra asserts added for the incremental GC, the original list validations from 3.13 are still present. |
pablogsal commentedNov 14, 2024
Sorry I was confused by the diff |
brandtbucher left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Looks good. I'm guessingvalidate_list is the cheapest (or most useful) check?
Uh oh!
There was an error while loading.Please reload this page.
The doctests CI takes ages and it uses a debug build.
Arguably debug builds should be good for debugging, i.e. lots of asserts. If you want performance don't use a debug build.
However, the incremental GC does adds alot of extra sanity checks and it is a pain if CI jobs take half an hour or more.
So, this PR adds an extra debug option to reduce the amount of checking done in a normal debug build, but keeps the extra sanity checks if needed for debugging the GC.
📚 Documentation preview 📚:https://cpython-previews--126777.org.readthedocs.build/