Commit Graph
13 Commits
Author SHA1 Message Date
Brian VaughnandGitHub b41b7f2e0b DevTools: Timeline profiler refactor
Refactor DevTools to record Timeline data (in memory) while profiling. Updated the Profiler UI to import/export Timeline data along with legacy profiler data.

Relates to issue #22529
2022-01-28 13:09:28 -05:00
Brian VaughnandGitHub 60228a0f7e Remove unstable scheduler/tracing API (#20037) 2021-04-26 19:16:18 -04:00
Brian VaughnandGitHub 15fed0570a Update DevTools Error strings to support GitHub fuzzy search (#21314) 2021-04-19 13:05:28 -04:00
Brian VaughnandGitHub fdc7b79ef2 DevTools shows which fibers scheduled the current update (#21171) 2021-04-09 10:35:06 -04:00
Brian VaughnandGitHub 0a5bf1a801 DevTools Profiler: Add commit and post-commit durations to UI (#20984) 2021-03-17 12:27:37 -04:00
49467f6862 Profiler: Skip reading element for imported data (#18913)
* skip reading element for imported data

* rename nodes & enable store lookup for components tab

* replace names

* Added some more test coverage; reverted rename

Co-authored-by: Brian Vaughn <[email protected]>
2020-05-15 10:38:01 -07:00
Brian VaughnandGitHub b52af9eb12 Disable Profiler commit filtering (#18862)
* Disable Profiler commit filtering

We used to filter "empty" DevTools commits, but it was error prone (see #18798). A commit may appear to be empty (no actual durations) because of component filters, but filtering these empty commits causes interaction commit indices to be off by N. This not only corrupts the resulting data, but also potentially causes runtime errors.

For that matter, hiding "empty" commits might cause confusion too. A commit *did happen* even if none of the components the Profiler is showing were involved.

* Restart flaky CI
2020-05-07 17:07:20 -07:00
cutjavascriptandGitHub 1520f7b0c4 dataForRoots.set duplicate removal (#17993)
dataForRoots.set duplicate removal
2020-02-06 11:48:02 -08:00
Brian VaughnandGitHub 137ecef6d1 DevTools Profiler: Improve how empty commits are filtered (#17771)
Previously, DevTools filtered empty commits on the backend, while profiling, through the use of a bailout heuristic that React currently happens to use. This approach was brittle and may have exacerbated the long-standing Profiler bug #16446.

This PR removes that heuristic and adds as a post-processing filtering pass instead. This removes the coupling between DevTools and a React internal implementation detail that may change.

I believe DevTools has two choices of criteria for this filtering:
* Filter commits that have no actual duration metadata.
* Filter commits that have no recorded operations (no mutations to the tree, no changed tree base durations).

I chose the first option, filtering by commits that have no reported metadata. It will miss an edge case, e.g. , but we would have nothing meaningful to show in the Profiler for those cases anyway. (This particular edge case is why one of the snapshots changed with this commit.)

The second option, filtering by recorded operations, could potentially miss a more important edge case: where a component *did* render, but its  didn't change. (In that event, there would be no operations to send.)
2020-01-04 16:50:39 -08:00
Brian Vaughn a0b66f405d Added FB copyright header 2019-08-27 10:54:01 -07:00
Brian Vaughn 73aec5cc57 Prettier 2019-08-13 17:58:03 -07:00
Brian Vaughn 0da4f3f4aa Reorganized folders into packages/* 2019-08-13 15:59:43 -07:00
Brian Vaughn 1435d5c8d3 Reorganized things again into packages 2019-08-13 11:37:25 -07:00