Update with upstream june 2026

[chill389cc]

WIP. I think I scrapped some of Jareds sorting changes because I didn't want to deal with them for initial branch creation, but I'll go back in to see if I can re-implement them before I try to merge this.

Created branch update-with-upstream-june-2026 and rebased it onto upstream/master (your remote is master, not main). The branch is now 0 behind upstream, with 4 commits of our unique work on top:

24586d0 publish a compiled binary as a release on push
b728e61 Add a toggle for compact mode
11cb725 build deps for x86 and arm
3279eb8 Allow sorting by fetch and report stages

What happened to the 12 commits

- 6 merge commits dropped automatically; they carried no unique content.
- 222b62e "Only track execute workers" auto-dropped by git (identical patch-id to upstream's 1c993ed; truly already upstream).
- 31ab862 "Display ReportResultStage slots"  skipped. This is the one you flagged as "merged with minor changes": upstream's 4a1f669 is the same commit by the same author, better integrated into their refactored renderWorkerRow (which now has a view parameter). Ours added nothing theirs lacked, so keeping it would have clobbered their refactor.
- 4 genuine commits replayed and kept (above).

Conflicts resolved

- view/queue.go (sorting): ported our multi-stage sort (Name/Fetches/Executions/Reports) into upstream's new sortWorkers function, keeping upstream's *profileResult types and richer name comparator.
- view/queue.go (compact toggle): the commit predated upstream's view refactor, so I surgically re-applied just the compact feature � compact field, c keybind, and the compact/expanded bar-width logic for all three stage bars � onto the current code rather than hand-merging the giant blocks.
- BUILD / MODULE.bazel: kept our x86/arm binary, tar, and deb targets plus the rules_pkg dep.
- MODULE.bazel.lock: took upstream's generated lockfile as the base and let bazel regenerate it (hand-merging a generated 18k-line lock would be error-prone).

We'll wait for George to fix the SIGSEGV that I get when even just using buildfarm/bf-client's latest on our cluster. THen update this again, and then force-push to 'merge'

[chill389cc]

Right now this is setup so that, when approved, we force-push to lucid and then close the PR. We could recreate the PR to be based on lucid instead of the common ancestor between lucid and upstream, but this produces a cleaner commit history. Just a toss-up I guess.