[axisrow]

Summary

Mirror of the upstream fix PR: sirmalloc#739 (issue: sirmalloc#738, moved from #62).

checkPowerlineFonts() missed Nerd Fonts v3 file names (no spaces, e.g. JetBrainsMonoNerdFont-Regular.ttf), the short NF form, and fonts nested in subfolders. Fix: /nerd ?font/i + an NF token in the patterns, recursive depth-capped scan of the font folders (extracted into the exported fontFolderHasPowerlineFont for testing against real temp dirs).

The branch is cut from upstream/main, so the diff also carries the upstream commits the fork does not have yet (up to bed8501); the powerline change itself is the single fix(powerline): … commit.

Test plan

• src/utils/__tests__/powerline-font-detection.test.ts — the scanner against real temp directories; on pre-fix code it fails to import.
• bun run lint clean.
• Upstream CI green on fix(powerline): detect Nerd Fonts v3 names and fonts in font subfolders sirmalloc/ccstatusline#739.

🤖 Generated with Claude Code

[axisrow]

Approving. Reviewed the powerline fix commit d621a3b: the extraction into exported fontFolderHasPowerlineFont (depth cap, hidden-entry skip, per-directory error swallowing) preserves the old semantics while adding recursion, the async path reuses the same code so no duplicated pattern list is left behind, and the new tests on real temp dirs cover v3 spaceless names, the short NF form, subfolder nesting, the Debian opentype layout, the old spaced v2 name, and the negative cases. The rest of the diff is the upstream sync the branch was cut with, as disclosed in the description; the powerline change and its test are byte-identical to upstream sirmalloc#739 head, and merge 1463381 keeps the fix intact against fork main. No required changes. Optional, non-blocking: (1) /nf(?=[-.0-9]|$)/i is case-insensitive with no left boundary, so any filename whose tail is nf (a stray *.conf inside a font folder) counts as a Nerd Font; risk is low for a heuristic check, but the v3 short form is always uppercase, so a case-sensitive /NF(?=[-.0-9]|$)/ would still match JetBrainsMonoNF-Regular.ttf while trimming that; note the spaced short form (MesloLGS NF Regular.ttf) is not matched by the lookahead either way, add \s to the lookahead if that naming matters. (2) The test helper could use dirname(filePath) instead of join(filePath, ..); recursive mkdir resolves both, dirname is just clearer. (3) Symlinked font directories are not followed (Dirent.isDirectory is false for symlinks); not a regression, and the fc-list fallback covers fontconfig-registered setups.

[axisrow]

Approving. Re-review of the updated head: since the previous review (5527921612, approved) the branch gained only two empty retrigger-CI commits (6cc6da6, e1e0a35); the tree is byte-identical (diff against the previously reviewed 1463381 is empty), so the powerline fix and its tests are unchanged. Verdict stands: the extraction into exported fontFolderHasPowerlineFont (depth cap, hidden-entry skip, per-directory error swallowing) preserves the old semantics while adding recursion, the async path reuses the same code, and the temp-dir tests cover the v3 spaceless names, the short NF form, subfolder nesting, the Debian opentype layout, the old spaced v2 name, and the negative cases. No required changes. Optional, non-blocking (carried over from the previous review): (1) /nf(?=[-.0-9]|$)/i is case-insensitive with no left boundary, so any filename whose tail is nf (a stray *.conf inside a font folder) counts as a Nerd Font; a case-sensitive /NF(?=[-.0-9]|$)/ would still match JetBrainsMonoNF-Regular.ttf while trimming that, and the spaced short form (MesloLGS NF Regular.ttf) is not matched by the lookahead either way. (2) The test helper could use dirname(filePath) instead of join(filePath, ..). (3) Symlinked font directories are not followed; not a regression, and the fc-list fallback covers fontconfig-registered setups.

[axisrow]

Approving. This run targets the same head e1e0a35 already reviewed and approved in review 5527982924: the tree is unchanged, base a2c1161 unchanged, and no new upstream feedback has appeared on sirmalloc#739. Verdict stands. The powerline fix extracts the scan into exported fontFolderHasPowerlineFont (depth cap, hidden-entry skip, per-directory error swallowing) while preserving old semantics, the async path reuses the same code, and the temp-dir tests cover the v3 spaceless names, the short NF form, subfolder nesting, the Debian opentype layout, the old spaced v2 name, and the negative cases. No required changes. Optional, non-blocking (carried over): (1) /nf(?=[-.0-9]|$)/i is case-insensitive with no left boundary, so any filename whose tail is nf (a stray *.conf inside a font folder) counts as a Nerd Font; a case-sensitive /NF(?=[-.0-9]|$)/ would still match JetBrainsMonoNF-Regular.ttf while trimming that, and the spaced short form (MesloLGS NF Regular.ttf) is not matched by the lookahead either way. (2) The test helper could use dirname(filePath) instead of join(filePath, ..). (3) Symlinked font directories are not followed; not a regression, and the fc-list fallback covers fontconfig-registered setups.

[axisrow]

Approving. This run targets head 5747b82, which adds only an empty retrigger-CI commit on top of the previously approved e1e0a35 (review 5527997048): the tree diff between the two is empty, base a2c1161 unchanged, so the powerline fix and its tests are byte-identical to what was already approved. Verdict stands. The fix extracts the font scan into exported fontFolderHasPowerlineFont (depth cap, hidden-entry skip, per-directory error swallowing) while preserving old semantics, the async path reuses the same code, and the temp-dir tests cover the v3 spaceless names, the short NF form, subfolder nesting, the Debian opentype layout, the old spaced v2 name, and the negative cases. No required changes. Optional, non-blocking (carried over): (1) /nf(?=[-.0-9]|$)/i is case-insensitive with no left boundary, so any filename whose tail is nf (a stray *.conf inside a font folder) counts as a Nerd Font; a case-sensitive /NF(?=[-.0-9]|$)/ would still match JetBrainsMonoNF-Regular.ttf while trimming that, and the spaced short form (MesloLGS NF Regular.ttf) is not matched by the lookahead either way. (2) The test helper could use dirname(filePath) instead of join(filePath, ..). (3) Symlinked font directories are not followed; not a regression, and the fc-list fallback covers fontconfig-registered setups.

[axisrow]

Approving. The new commit 77af91b fixes three merge fallouts against origin/main, and each checks out against the code on this head: (1) the terminal-sanitize expectation okx matches the merged upstream c08a59d behavior, where the 8-bit CSI is consumed as one complete sequence (skipEscapeSequence sees the 0x9b introducer, 2 as a parameter byte, J as the 0x40-0x7e final byte, and drops the whole run); (2) the usage-prefetch spend-limit assertion now uses anyCacheMatcher, which is defined in the file (line 29) and matches every neighboring assertion for the fork cache-handle argument; (3) the 5h: compact preset matches the actual label the session usage widget now renders (usage-percent-widget config label 5h:), and keeping Session: alongside it is correct because SessionClock still renders that prefix. CI on this head is green (Build, Lint & Type Check, Test, both runs). The powerline fix itself is unchanged since review 5528674718. No required changes. Optional, non-blocking: 5h: -> 5h: only trims the trailing space, the point is keeping the widget covered by the compact-label probe, same as the existing Mem: -> Mem: entry; fine as is.