Import frame line numbers from Chrome profiles#6192
Open
canova wants to merge 3 commits into
Open
Conversation
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #6192 +/- ##
==========================================
+ Coverage 83.63% 83.66% +0.02%
==========================================
Files 346 346
Lines 37174 37248 +74
Branches 10318 10441 +123
==========================================
+ Hits 31090 31163 +73
- Misses 5656 5657 +1
Partials 428 428 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
The Chrome/Node importer derived each frame's index from `nodeIndex - 1` and wrote into the shared global frame table. Because V8 node ids restart at 1 for every thread, two threads' node 1 both mapped to frame 0, so the later thread clobbered the earlier thread's frame data (func, category, line) while the earlier thread's stacks still pointed at it. Allocate a fresh frame index from `frameTable.length` for each node instead. This is also a prerequisite for emitting more than one frame per node.
V8 cpuprofiles carry two kinds of source locations per node: callFrame contains the function definition location, while positionTicks is the per-line breakdown of where the function self time was spent. We only imported the former, so the source view attributed all of a function self time to its definition line. Keep the definition location in the func table and use a neutral structural frame for the node so descendant samples do not inherit an arbitrary executed line. Create a sibling frame for every distinct valid positionTicks line and distribute only the node leaf samples among them in proportion to the tick counts using deterministic largest-deficit rounding. Leave their columns null because positionTicks does not contain column data, and ignore positionTicks on native frames without a source. positionTicks lines are 1-based, unlike the 0-based callFrame line number.
The DevTools trace-event ProfileChunk format records the executed source line and column for each sample in args.data.lines and args.data.columns, parallel to cpuProfile.samples, rather than the per-node positionTicks of the .cpuprofile format. Read these arrays and put each emitted sample to a frame carrying its executed line. The frame is created lazily as a sibling of the node base stack. Native frames without a source fall back to the base stack, and the positionTicks path stays as a fallback when no per-sample line is present.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Main | Deploy preview
Fixes #6182
It isn't super easy to see what this does without having the source code in the source view. But you can check the frame table to see that we have line and column numbers now.
For example, you can use this chrome trace: Trace-20260715T220108.json.gz
Check the frameTable in the before and after profiles with:
It'll look like now we have less information in the frame table, but no. Previously, we were always adding the line and column of the function definition, which wasn't correct. Now these numbers are in the funcTable. And we only have line and column number if the frame truly has line/column. Otherwise we omit that information.