self-profiling: Add events for everything except trait selection.

[michaelwoerister]

This is the followup PR to #64840.

Trait selection events are still missing (at least those not covered by regular queries).

r? @wesleywiser (or @Mark-Simulacrum if @wesleywiser is not available at the moment)

[Mark-Simulacrum]

Just to make sure -- this won't run into the problem we hit earlier with non-contained events, right? I think the answer is no, because all added events are scope-specific. But I'd like to ask to make sure.

r=me modulo that concern

[Mark-Simulacrum]

I could be wrong -- but I think we already have events with spaces in them, in which case this can probably be nicer if we just use spaces directly (both in this event and generally).

[wesleywiser]

Yeah, we do. Unless there's a strong reason to use underscores (for instance, if this activity was just the name of the enclosing function), I'd recommend using spaces as well.

[michaelwoerister]

I'm not sure. In my opinion it's more like: unless there is a strong reason for doing otherwise, it's better to keep any kind of identifiers as simple as possible (e.g. allowing only things that would also make valid Rust identifiers). Otherwise one might end up with headaches down the road, like having to weirdly escape things in commandline strings, or not knowing if trailing whitespace is significant or not. The identifiers emitted here will go through various processing tools (many of which we don't now about yet). I'd rather keep things simple.

[wesleywiser]

Seems reasonable to me.

[wesleywiser]

cgu partitioning: place roots perhaps?

[bors]

☔ The latest upstream changes (presumably #65223) made this pull request unmergeable. Please resolve the merge conflicts.

[wesleywiser]

@bors r+

[bors]

📌 Commit 38e0b5c15d4195be42298cb361316c4c350da957 has been approved by wesleywiser

[wesleywiser]

Woops... missed the merge conflict message.

@bors r-

r=me with merge resolved

[michaelwoerister]

Just to make sure -- this won't run into the problem we hit earlier with non-contained events, right? I think the answer is no, because all added events are scope-specific. But I'd like to ask to make sure.

Yes, the RAII-based API should make sure that something like that can't happen.

[michaelwoerister]

@bors r=wesleywiser

[bors]

📌 Commit ceb1a9c has been approved by wesleywiser

[michaelwoerister]

Thanks for the review!

[bors]

⌛ Testing commit ceb1a9c with merge 321ccbe...

[bors]

☀️ Test successful - checks-azure
Approved by: wesleywiser
Pushing 321ccbe to master...