Skip to content

Bump alpine from 3.20 to 3.24 - #11

Closed
dependabot[bot] wants to merge 1 commit into
masterfrom
dependabot/docker/alpine-3.24
Closed

dependabot[bot] wants to merge 1 commit into
masterfrom
dependabot/docker/alpine-3.24

Conversation

@dependabot

@dependabot dependabot Bot commented on behalf of github Aug 30, 2026 •

Copy link
Copy Markdown

Warning

Cooldown could not be applied because no publication date was available from the registry.

Bumps alpine from 3.20 to 3.24.

@dependabot dependabot Bot added dependencies Pull requests that update a dependency file docker Pull requests that update docker code labels Aug 30, 2026
@dependabot
dependabot Bot force-pushed the dependabot/docker/alpine-3.24 branch 2 times, most recently from 67284a5 to 66709fa Compare September 1, 2026 16:39
Bumps alpine from 3.20 to 3.24.

---
updated-dependencies:
- dependency-name: alpine
  dependency-version: '3.24'
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot
dependabot Bot force-pushed the dependabot/docker/alpine-3.24 branch from 66709fa to 95ada77 Compare September 4, 2026 15:13
@dependabot @github

dependabot Bot commented on behalf of github Sep 4, 2026

Copy link
Copy Markdown
Author

Superseded by #14.

@dependabot dependabot Bot closed this Sep 4, 2026
@dependabot
dependabot Bot deleted the dependabot/docker/alpine-3.24 branch September 4, 2026 15:55
karthik1729 added a commit that referenced this pull request Sep 12, 2026
…, and stop what a generation bump left running

Findings #28, #33, #35, #37, plus stop.rs #9, mounts #10, intercept port #11
and wishes index #12.

- `pull_one` bounds the COPY into `btrfs receive` and the child's own exit by
  `peer_send_timeout_secs`, killing it on expiry. The request timeout could
  only fire on a body still being read, so a receive that stopped reading — or
  that never exited after EOF — held the volume's pull slot for the life of
  the process.
- The janitor's `btrfs subvolume delete` runs under `timeout -s KILL 300`: it
  is a blocking call on the janitor's own thread and `btrfs` has no deadline
  of its own, so one wedged delete stopped every other sweep.
- The peer send handler's Volume GET is bounded at 10s and an unreadable
  answer is UNKNOWN, not `quota_gb: 0` — refusing a send on a guessed quota
  costs a replica its only source, and the puller's own `take(max + 1)` still
  bounds what lands.
- `mount_homes` attempts one repair per 30s window and tells a concurrent
  caller so rather than queueing behind a mount against a dead export.
- `stop_environment` runs the teardown and recomputes `Replicated` before
  stamping `observed_generation` on a stop at an older generation.
- `mkdir_env_mounts` validates every mount before creating any: dedup is by
  folder, so a second service's invalid mount was never checked.
- `intercept_plan` refuses a port rewrite naming a port the service does not
  declare (and 0 on either side), settling the wish `Off/PortsInvalid`; the
  slice matches by port name, so such a rewrite scaled the real service to
  zero and delivered the traffic nowhere. `plan`/`wishes` are read with `get`,
  never indexed — a missing key in a reconciler is a downed controller.

Tests: the 413 test now serves a real Volume (it had been asserting the
refusal that an unreadable one produced), with its sibling for the unknown
case; unit tests for the mount dedup and the port-map check; a reconcile test
for the generation bump while stopped.
karthik1729 added a commit that referenced this pull request Sep 12, 2026
Review findings #6, #7, #11, #15, #16, #18, #80, #81, #129.

Every fix here is one shape: a read that failed was being turned into a
value somebody then acted on. A stat that errored listed an image as
empty; a visibility read that errored flipped a public repo out of every
listing; a body read that errored told the pusher their pack had holes.
They now skip, and say what happened.

Alongside them: poisoned mutexes recover instead of wedging the node
(`LockOrRecover`), accepted peer streams get the dead-peer bounds the
dialling side already had, the stranded-merge lane paces whatever its
scan found, `remove_protection` validates its pattern like `set` does,
a marker's description can no longer forge the fields after it, a blank
or too-short `KLOUDLITE_JWT_SECRET` refuses the boot by name instead of
panicking, and `head_target` has no `unwrap`.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency file docker Pull requests that update docker code

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants