Noticed while checking citations for the rest of this sweep. Class: routine post-burst coverage lag on the manual-verification checklist, plus one entry that now names the wrong GTK call.
Coverage
docs/live-verify.md:17 declares "Coverage: #458 through #634". Merged since f7a980d (#636, the last refresh): ee96105 (#637), 743e106 (#639), cacd997 (#642).
#637 and #639 each carry a substantive "## Live-verify note". #637's asks for a glib::WeakRef / gtk::Window::toplevels() count check across repeated kanshi hot-plug cycles for never-opened layer windows — a real un-CI-testable gesture that currently exists nowhere but a merged PR body, which is exactly the loss this file was created to prevent (#611's own note: "this file goes stale within minutes of a merge burst, so updating it belongs to the burst rather than to a later pass").
The #630 entry names a call the code no longer makes
docs/live-verify.md:359-361 tells the verifier the fix means close_all "no longer holds the PANELS RefCell borrow across each window.close() call". trollshell/src/modal.rs:1008 now reads panel.window.destroy();, changed by #637, whose comment at modal.rs:1000-1003 explains why close() was wrong. The entry is phrased in the present tense as what the current code does, so a reader diffing it against modal.rs sees a mismatch.
What I checked and dropped
I expected #642 (the corrected default station id) to need a caveat, since crates/hytte-services/src/places.rs:341 writes the default config only when places.toml is absent — which would make the departures steps at docs/live-verify.md:160/:193 read as a false pass against an existing config. That is already handled outside this file: the #641 triage comment says so verbatim ("Your existing file won't be fixed by that … You'll need to edit ~/.config/trollshell/places.toml yourself") and #640 is open on the missing editor UI. No doc change is needed for it beyond the coverage line.
Severity: low. Three PRs of lag on a file the maintainer folds in by habit, plus a one-word correction. It is worth writing down only because #637's toplevel-count gesture is recorded nowhere else.
Recommendation
Fold #637/#639/#642 in and bump the coverage line at :17 to #642, carrying #637's toplevels()-count gesture across verbatim. While there, change close() to destroy() at docs/live-verify.md:359-361. Good rider on the next docs PR; not worth a standalone one.
Filed by the 2026-07-30 residual-work sweep (index: see the meta issue). Severity assessed low. Every citation was re-verified against HEAD before filing.
Noticed while checking citations for the rest of this sweep. Class: routine post-burst coverage lag on the manual-verification checklist, plus one entry that now names the wrong GTK call.
Coverage
docs/live-verify.md:17declares "Coverage: #458 through #634". Merged since f7a980d (#636, the last refresh): ee96105 (#637), 743e106 (#639), cacd997 (#642).#637 and #639 each carry a substantive "## Live-verify note". #637's asks for a
glib::WeakRef/gtk::Window::toplevels()count check across repeated kanshi hot-plug cycles for never-opened layer windows — a real un-CI-testable gesture that currently exists nowhere but a merged PR body, which is exactly the loss this file was created to prevent (#611's own note: "this file goes stale within minutes of a merge burst, so updating it belongs to the burst rather than to a later pass").The #630 entry names a call the code no longer makes
docs/live-verify.md:359-361tells the verifier the fix meansclose_all"no longer holds thePANELSRefCellborrow across eachwindow.close()call".trollshell/src/modal.rs:1008now readspanel.window.destroy();, changed by #637, whose comment atmodal.rs:1000-1003explains whyclose()was wrong. The entry is phrased in the present tense as what the current code does, so a reader diffing it againstmodal.rssees a mismatch.What I checked and dropped
I expected #642 (the corrected default station id) to need a caveat, since
crates/hytte-services/src/places.rs:341writes the default config only whenplaces.tomlis absent — which would make the departures steps atdocs/live-verify.md:160/:193read as a false pass against an existing config. That is already handled outside this file: the #641 triage comment says so verbatim ("Your existing file won't be fixed by that … You'll need to edit~/.config/trollshell/places.tomlyourself") and #640 is open on the missing editor UI. No doc change is needed for it beyond the coverage line.Severity: low. Three PRs of lag on a file the maintainer folds in by habit, plus a one-word correction. It is worth writing down only because #637's toplevel-count gesture is recorded nowhere else.
Recommendation
Fold #637/#639/#642 in and bump the coverage line at
:17to #642, carrying #637'stoplevels()-count gesture across verbatim. While there, changeclose()todestroy()atdocs/live-verify.md:359-361. Good rider on the next docs PR; not worth a standalone one.Filed by the 2026-07-30 residual-work sweep (index: see the meta issue). Severity assessed low. Every citation was re-verified against
HEADbefore filing.