Skip to content

release: 6.3.4 - #4

Merged
vibesoftwarecoder merged 1 commit into
masterfrom
release/6.3.4
Sep 10, 2026
Merged

release: 6.3.4#4
vibesoftwarecoder merged 1 commit into
masterfrom
release/6.3.4

Conversation

@vibesoftwarecoder

Copy link
Copy Markdown
Owner

Cuts 6.3.4. app/version.txt 6.3.3 → 6.3.4 plus notes. The tag must equal v<version.txt> or the release build fails on purpose, so the bump comes first.

No code changes — ships two already-merged commits:

⛔ The discovery change has never been run

It compiles under Qt 5.14.1 and Qt 6, and the host side is confirmed — a seat really does answer /serverinfo with the name discovery looks for. But no client has executed this code. The machine it was written on has no Qt toolchain, so a build could not be tested there.

The plausible failure is client-side: discovery reads the host list every 15 seconds while other paths write to it under a write lock on a thread pool. A pull model (aboutToPoll) was chosen specifically to avoid deadlocking against that — but designed-around is not tested.

The release notes lead with this rather than bury it, name the symptom to watch for (a hang on startup or when adding a host), and point at 6.3.3 as the way back. Understating it would be the wrong trade for a client binary people install.

What users get if it works

Seats appear on their own from any machine, within about fifteen seconds of being provisioned, with no API key and no need to expose MultiSeat's dashboard to the network. Before this they only ever appeared if Moonlight was running on the streaming host.

The manual fallback is documented in the notes: host address plus the seat's port, starting at 48100 and rising by 30 per seat. That is also the answer for a host with non-default port settings, since discovery checks the defaults and nothing advertises a custom range.

Once merged, pushing v6.3.4 publishes both the Windows and macOS assets.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SQvL62WkT8xDWXqyjFCGDw

app/version.txt 6.3.3 -> 6.3.4, plus notes. The tag must equal v<version.txt>
or the release build fails on purpose, so the bump comes first.

Ships two merged commits: 00e256a (seats are found by probing the seat port
block instead of asking the MultiSeat service at 127.0.0.1, which could only
ever work with Moonlight running on the host) and 1c5f539 (the PR build
pipeline trimmed to the jobs that pass).

⛔ The notes lead with the fact that the discovery change has NEVER BEEN RUN.
It compiles under Qt 5.14.1 and Qt 6, and the host side is confirmed -- a seat
answers /serverinfo with the name discovery looks for -- but no client has
executed this code. The machine it was written on has no Qt toolchain.

The plausible failure is client-side: discovery reads the host list every 15s
while other paths write to it under a write lock on a thread pool. A pull model
was chosen specifically to avoid deadlocking against that, but designed-around
is not tested. The notes say so, name the symptom to watch for, and point at
6.3.3 as the way back.

Understating that would be the wrong trade for a client binary people install.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SQvL62WkT8xDWXqyjFCGDw
@vibesoftwarecoder
vibesoftwarecoder merged commit 2db0d0e into master Sep 10, 2026
2 checks passed
@vibesoftwarecoder
vibesoftwarecoder deleted the release/6.3.4 branch September 10, 2026 15:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant