diff options
| author | user <user@clank> | 2026-08-27 16:40:08 +0200 |
|---|---|---|
| committer | user <user@clank> | 2026-08-27 16:40:08 +0200 |
| commit | e0abf6b0b98e17cf7cc8e8a81358a6b18ae52f8c (patch) | |
| tree | fe55474ca9263dd4a8daa82f7390c496d1083559 /src/expr.c | |
| parent | pty: open the slave before the fork, so a pane's size reaches it (diff) | |
| download | slosh-e0abf6b0b98e17cf7cc8e8a81358a6b18ae52f8c.tar.gz slosh-e0abf6b0b98e17cf7cc8e8a81358a6b18ae52f8c.tar.xz slosh-e0abf6b0b98e17cf7cc8e8a81358a6b18ae52f8c.zip | |
macos-remote-build: a synced tree must not read as dirty
web/ and tests/fuzz/ are excluded from the sync (hundreds of megabytes,
none of it an input to 'make all') but both are tracked, and .git goes
over whole -- so the moment a commit touched web/, the builder's stale
copy read as a local edit and git describe stamped a clean release tree
-dirty. slosh-0.1.1-dirty-macos-arm64.zip was signed, notarized and
correctly labelled as a tree nobody had: the exclusion's debris, not
the release.
After the rsync, restore the excluded tracked paths from the .git that
just went over: no network, no bandwidth lost to the exclusion, and a
tree that is really dirty still says so.
Diffstat (limited to 'src/expr.c')
0 files changed, 0 insertions, 0 deletions
