gh-120144: Disable the CALL event when possible to achieve zero overhead pdb #131390
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
iritkatriel
approved these changes
Mar 18, 2025
seehwan
pushed a commit
to seehwan/cpython
that referenced
this pull request
Apr 16, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is one line of well-thought code :)
We basically solved the debugger overhead in a big loop by disabling LINE events, the one (major) thing left was recursive calls - or rather, calls in general. We won't prevent CALL events from happening for every unrelated call because that would require a much larger change, but we can disable CALL events after we hit it once.
Notice that CALL event has a two effects which made it a bit more complicated:
step)No 2. makes it a bit challenging and much more rewarding.
We need to make sure we don't need any new events on this code object, before we can disable the CALL event.
First of all, if there's any breakpoint on the code object, we can't do that. But we already have check for that where the code was added.
Then we'll realize, if it's a
next,untilorreturn- it will just work fine, because they all rely on events on existing frame and code objects. It's impossible for those commands to stop at a new frame.The only command that could stop at a new frame, is
step- but if it's a step, it would eitherstop_here()already on this specificCALLevent so we can't reach to this pointskipargument, which means it won't stop in this function in the futureSo really, we don't need to care about anything. If we don't need to trace this function now, we won't need any new events for this code object in the future (until the user interaction), so we can simply disable the CALL event.
This basically gives us a zero overhead debugger. There's no observable overhead for the recursive fib implementation.
sys.monitoringfor pdb/bdb #120144