| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
flock(2) on the lock file's descriptor replaces "write my pid, and
decide whether the pid in the file is still alive". The kernel drops
the lock when the process ends, however it ends, so there is no stale
krino lock to detect and no takeover to race over.
What that fixes:
- Two runs that both judged a lock stale could remove and recreate it
and both believe they held it. Remove-then-create cannot be made
atomic; there is nothing to make atomic now. Pinned by a test with
eight callers over twenty rounds.
- A pid reused after a crash made the lock live for ever, and the
message named neither the file nor the pid, so there was nothing to
act on. The message now names both.
- Signal(0) reads EPERM as "not running", so a lock held by another
user was taken over. There is no such judgement left to get wrong.
A run that waits for a held lock now says so first. Waiting is what
the spec asks for, but the wait has no timeout, and in silence it is
indistinguishable from a hang - I spent two minutes on one myself
today, waiting on a lock the window was holding.
Tests that faked a held lock by writing a file now hold a real one.
|
| |
|
|
|
|
|
|
|
|
|
| |
So that go install can find it. cgit serves no go-import meta tag, so
nginx answers a ?go-get=1 request for /NAME with one pointing at
https://git.labunix.xyz/NAME.git; the alternative was carrying a .git
suffix through every import line forever.
One sed over the imports, both go.mod files, and the dependency gate's
whitelist. Nothing else depends on the path. go install works from the
next tag, the first release whose go.mod carries it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
About 340 comments cited the development process: task and plan
numbers, fix waves, rulings, reviewers, and the author in the third
person with a date. None of that exists outside the work itself, so to
a reader it pointed at nothing. Each one now states the engineering
reason it was standing in front of; where a comment was provenance and
nothing else, it is gone.
References to docs/design.md and docs/gui-design.md by section stay:
both ship with the repository. The design documents lose their
amendment diaries - CHANGELOG.md is that record - and the GUI's says
plainly that the window has gone further than the document.
Only comments changed. Every .go file was parsed and its code printed
with comments stripped, before and after: the two hashes are identical
across all 175 files.
|
| |
|
|
| |
cleanup, text turning binary, explain flags, interrupt docs
|
| |
|
|
| |
w after error)
|
| | |
|
| |
|
|
| |
names, conflict enum, dependency gate, absolute tool paths
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
| |
One palette type styles the plan table, warnings, headers, outcome counts,
prompt keys, undo's refused steps and krino log's (undone), from the
16-colour ANSI palette plus bold and faint only. Widths are measured on the
plain text, so columns line up; with colour off the output is unchanged.
--no-color works before or after any subcommand, as NO_COLOR does. The two
search-and-replace colourings are gone.
|
| |
|
|
|
|
|
|
|
| |
README, changelog
Also: undo removes the directories its run created; a hardlink is never a
duplicate of its own other name; a flag written before "undo" is honoured;
--version prints no leading v. Duplicate conditions with different scopes
not sharing an original is documented as a known limitation.
|
| |
|
