Instruments importer #3684
Instruments importer #3684
Conversation
This commit should not be reviewed as code written by the author, but rather as code that is imported pretty much directly. Only minimal changes were made like removing export declarations, or import syntax. This inlines the code from several different files as much as possible.
This code is useful if we want to support this feature, but for now, just comment it out so that Flow doesn't complain.
This commit represents trivial lint and style fixes on the code. No behavior was changed here.
This is the first real new code. It changes the implementation from using the non-standard webkitEntries API, the JSZip API.
|
I filed jlfwong/speedscope#384 for an Instruments file that wouldn't load with this and speedscope. |
|
In order to land this PR, the first few commits are all importing speedscope code and making it pass some basic CI requirements. These shouldn't be too controversial. The two WIP commits on the end are the ones getting it to work for the Firefox Profiler. Honestly, they could probably be squashed into a single commit, and rebased, and landed as is. I'm not planning on finishing this PR at this time. There is an issue that @jrmuizel filed, which I believe that certain Instruments profiles have different file organization structure. I was testing this code from instruments-simple.trace.zip which appears to be working in the deploy preview. |
|
It seems like they changed the Instruments file format quite a bit after Xcode 10.
Interestingly the old Instruments profiles don't load in the new Instruments version as well. So it seems like they dropped the support for those profile types completely. |
This was an attempt to port the speedscope importer to the profiler. I think it was sort of working when I stopped.
I initially attempted to fix up the intern's code, but ran into issues with not knowing what was a bug in the new code, or if there were bugs in the original code.
My process was to naively copy over the relevant speedscope code, and then re-work it to get the types passing. I think there were some existing issues with how an instruments profile was recorded and the final structure of recorded profile. There is an assumption in this code that the profile was either recorded from the command line, or recorded from the instruments app. I don't recall which at this point.
┆Issue is synchronized with this Jira Task