| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The tile counts what can be installed now (Slackware and the binary
repo) and what only needs a build, and goes accent when there is
something to install. It runs updatesctl while the drawer is open and
not otherwise; a result over 30 minutes old is followed by a refresh
that runs nvchecker, never after a refresh, so a failing nvchecker
cannot loop.
The page has one section per next action: Slackware upgrades, new and
removed first, then upgrade here, awaiting build, upstream newer, app
updates, the order the work is done in. Each row names the package, the
repo it comes from and the one version jump its action covers. A source
that could not be read says so under the section it empties instead of
reading as nothing pending.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A git hash starting with a letter (3.1-a59215d) ordered behind its base,
because any trailing alphabetic chunk counted as a pre-release; only
alpha, beta, rc, pre and dev do now. Trailing zeros (1.14 against
1.14.0) compare equal instead of reading as an upstream bump. A bad line
in the slackpkg blacklist is skipped rather than crashing the script and
leaving the drawer with nothing. An AppImage keeps its key over a
flatpak of the same name, as the spec orders them, and flatpak list has
a timeout. A Slackware upgrade compares version and build, so an arch
change alone is not a row. The spec's prose now matches the code: the
upstream comparison includes the installed version.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The awaiting-build branch also required the build to be older than
upstream, which cannot fail there: a build newer than installed was
already taken as an upgrade, so the build is at most the installed
version, which is behind upstream whenever a row is emitted. Removing
it left every test passing, which is how it was found.
The blacklisted testing/ entry sat after the main one in the fixture, so
setdefault kept the main one and the mirror-side blacklist was never
exercised; it now comes first, and dropping the check fails the test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A fifo with no writer blocks cat at open() the way a hard NFS mount
blocks during a NAS outage. The repo section reports unreachable within
the timeout and the mirror sections still fill.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
|
Compared mirror against installed, as upgrade-all does, not against
slackpkg's local ChangeLog: `slackpkg update` replaces that without
installing anything, so a diff against it reads clean after an update
that was never applied. New and removed come from the mirror's
ChangeLog, the newest Added or Removed line deciding, so a package
added and later upgraded still counts as new. Blacklisted build tags
and directories are skipped, as slackpkg skips them.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
|
Each nvchecker stanza lands in one section: upgrade here when the binary
repo's PACKAGES.TXT holds a newer build than installed, awaiting build
when the recipe (the user's or SBo's) has caught up but the build has
not, upstream newer when the recipe lags, app updates for AppImages and
flatpaks. Built comes from PACKAGES.TXT rather than the .info, which is
why gamescope and kitty no longer read as installable: the repo still
held the old build. The repo is on a hard NFS mount, so it is read by a
child the timeout can kill.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
|
nvtable compares versions for equality, which cannot tell an installed
2.1.278 from an SBo recipe of 2.1.267 that is older than it. vercmp
orders numeric chunks as numbers and treats a trailing alphabetic chunk
as a pre-release, so a kernel suffix or git hash orders as newer than
its base and never reads as behind.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
