Add per-builder build events to IResourceChangeEvent

[vogella]

Summary

Today the resources framework only fires workspace-level PRE_BUILD and POST_BUILD events.
Listeners learn that a build ran and get the aggregated delta, but they cannot see which builder ran, on which project, for which build kind, or how long each individual builder took.

This proposes two new event types, PRE_PROJECT_BUILD and POST_PROJECT_BUILD, fired once per builder execution and nested inside the surrounding workspace PRE_BUILD/POST_BUILD pair.

Benefits:

• Per-builder profiling. Tools can attribute build time to a specific builder and project instead of a single opaque workspace total, making it easy to spot the slow builder in a large reactor.
• Monitoring and diagnostics tooling. The granularity enables build-monitor views and telemetry that show the real build sequence, which the aggregate events cannot support.

Performance impact

Negligible by default.

When no listener is registered, dispatch is a single hasListenerFor bitmask check per event with no allocation, so normal builds are unaffected.
The per-builder path also skips resource-delta computation entirely, so even with a listener attached the added work is cheap metadata dispatch (project, builder id, build kind) rather than tree/delta building.

Compatibility

The change is a binary-compatible API addition (new int constants plus a default method), so existing listeners are unaffected.

[vogella]

Example view which would use these events:

[merks]

I wonder if this is really such appropriate place for somehow gathering such information. It seems to me something more appropriate for logging that gathers exactly measurement details one desired rather than needing to add a listener to an API that sudden produces a potentially thousands of events where it currently produces only two. It seems nothing really needs the resource deltas for these events, which appears to be perhaps another indication that this is not really the appropriate API for this.

Perhaps something internal that builds a non-API internal data structure that might be used to display the information. The information does seem interesting/useful...

[vogella]

@merks I developed that a while ago for my patched Eclipse as I had multiple clients that had super long build cycles and no way to identify who is actually responsible for that. Some had "dead" builder still running and resulting in long build times others had custom builder which were not effective. I push my change and if you want you can have a look.

[merks]

It's not a question of whether the information is useful. The values in the view do appear useful. It's a question of whether changing this API for what appears to be purely and only for the purpose of gathering measurements/statics is really an appropriate fit.

It seems not actually such a good fit. One might instead maintain an internal data structure with really good details of exactly what you want to measure and to show in some view that is a friend of that internal representing.

(Profilers are also good for gathering details about what is taking so long, and then for doing something about it in the hotspots.)

[iloveeclipse]

I also have some objections here, I've commented on #2802

[iloveeclipse]

I believe more tracing is the right solution for the problem, not new API.

[vogella]

Thanks @merks and @iloveeclipse for the suggestion I look into tracing as an alternative.

This might take a bit longer, as I soon will vanish into summer vacation and because I use this in a similar version already since a while in my custom aggregator build so my personal pressure is not that high to rework that approach.

[vogella]

Btw, let me know if you want my example build monitor to easily identify slow builders, I can push it on top on #2802.

[iloveeclipse]

No, thanks, no time. I definitely see a value in such monitor, but it should use tracing infrastructure, not new API, see #2792 for example.