profiler-cli: add a thread list command - #6273
Conversation
The only thread inventory was profile info, which sorts by CPU and prints processes rather than a flat table. thread list prints handle, name, process, pid, CPU and marker count for every thread, following counter list, with --sort cpu|index|markers|name, --search and --limit. All rows print by default, as counter list does; a truncated default would reintroduce the discoverability problem profile info has. The marker count is the derived count thread markers reports, not the raw marker-table length -- deriving merges start/end pairs and adds jank markers, so the two differ. thread info reports the raw length, and so disagrees with both.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #6273 +/- ##
==========================================
+ Coverage 83.79% 83.81% +0.01%
==========================================
Files 350 350
Lines 37583 37635 +52
Branches 10459 10482 +23
==========================================
+ Hits 31492 31543 +51
- Misses 5664 5665 +1
Partials 427 427 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
|
For the same profile (https://share.firefox.dev/4goHeyH), the current The new This doesn't feel perfect to me (I think I would prefer the default sort order to be the thread's start time, so it's possible to find threads that existed at a some something else we are investigating occurred), but already seems better than having no thread list command at all, so if it's not broken, maybe we could do other things as incremental improvements to it. |
Main | Deploy preview
The only thread inventory was
profile info, which sorts by CPU and prints processes rather than a flat table.thread listprints handle, name, process, pid, CPU and marker count for every thread, followingcounter list, with --sort cpu|index|markers|name, --search and --limit. All rows print by default, as counter list does; a truncated default would reintroduce the discoverability problem profile info has.The marker count is the derived count thread markers reports, not the raw marker-table length -- deriving merges start/end pairs and adds jank markers, so the two differ. thread info reports the raw length, and so disagrees with both.