R5 Publish v1.0.0 release #64

Open
opened 2026-11-09 05:29:46 +00:00 by claude · 2 comments
Owner

Depends on: R1, R2, R3, R4, H2, H3, H4, H5

Cut the v1.0.0 tag and a Gitea release on this repo, with the tar.gz, the .deb, the dedicated server and SHA-256 checksums.

Acceptance

  • make check, make match and make soak all pass on the tagged commit.
  • Downloaded release artifacts verify against their checksums and run in a clean container.
  • Release notes state what is known not to work and what hasn't been tested on real hardware.
  • Do not cut the release if a gate has failed. Report what is missing instead.
**Depends on:** R1, R2, R3, R4, H2, H3, H4, H5 Cut the `v1.0.0` tag and a Gitea release on this repo, with the tar.gz, the `.deb`, the dedicated server and SHA-256 checksums. **Acceptance** - `make check`, `make match` and `make soak` all pass on the tagged commit. - Downloaded release artifacts verify against their checksums and run in a clean container. - Release notes state what is known not to work and what hasn't been tested on real hardware. - Do not cut the release if a gate has failed. Report what is missing instead.
claude added this to the v1.0 Release milestone 2026-11-09 05:29:46 +00:00
claude added the type/chorepriority/high labels 2026-11-09 05:29:46 +00:00
claude added this to the Tidewall v1.0 project 2026-11-09 15:00:10 +00:00
eric referenced this issue from a commit 2026-11-14 23:31:10 +00:00
Author
Owner

Release cut as a draft, because the binaries need one step from Eric.

Done

  • Version bumped to 1.0.0 (f776d59). Tag v1.0.0 (annotated) is on f5227b7 and pushed.
  • Gates on f5227b7, all passing:
    • make check: 258 scripts, 698 tests.
    • make match: the command log replays to the same hash.
    • make soak (24 clients, 3 processes, 2 matches, 60 ms ± 20 ms, 2% loss): 2,458 hashes compared, 0 differing. Server p99 tick 6.7 ms. Client and server memory within the allowance.
  • The first soak on f776d59 failed. Match 0 was decided during the drop, so the client leak check had one sample per process instead of two. Hashes and timing passed. That was a harness gap, filed and fixed as H17 (#86) in f5227b7. Then all three gates were rerun from scratch on the fixed commit.
  • make dist-check and make deb-check passed on the tagged commit. The .deb was checked in a clean Debian 13 container. VERSION in the tarball says commit f5227b731d.
  • Draft release has the release notes, including what doesn't work and that nothing ran on real GPU hardware, plus both .sha256 files.

Not done: acceptance item 2 (download, verify, run)
The Gitea tools only take uploads as base64 in the call itself, which doesn't work for a 30 MB tarball and a 22 MB .deb. I didn't take an API token from any config to upload some other way. So the binaries aren't attached yet, and I couldn't download them back.

Eric: the two files are on tower-plus under build/release-v1.0.0/ in the repo checkout. They're git-ignored, and later make dist runs won't overwrite them.

  • tidewall-1.0.0-linux-x86_64.tar.gz, sha256 5050c0bd…07c26c8
  • tidewall_1.0.0-1_amd64.deb, sha256 b9af9e8e…cb614aee

Attach both to the draft and publish it. The next scheduled run will then download them anonymously, check them against the .sha256 files and run them in a clean tidewall- container. Leave it as a draft if you want to wait for your R6 playtest first.

Leaving this open with needs-eric until the uploaded files are verified.

— tidewall-build

Release cut as a **draft**, because the binaries need one step from Eric. **Done** - Version bumped to 1.0.0 (f776d59). Tag `v1.0.0` (annotated) is on f5227b7 and pushed. - Gates on f5227b7, all passing: - `make check`: 258 scripts, 698 tests. - `make match`: the command log replays to the same hash. - `make soak` (24 clients, 3 processes, 2 matches, 60 ms ± 20 ms, 2% loss): 2,458 hashes compared, 0 differing. Server p99 tick 6.7 ms. Client and server memory within the allowance. - The first soak on f776d59 failed. Match 0 was decided during the drop, so the client leak check had one sample per process instead of two. Hashes and timing passed. That was a harness gap, filed and fixed as H17 (#86) in f5227b7. Then all three gates were rerun from scratch on the fixed commit. - `make dist-check` and `make deb-check` passed on the tagged commit. The .deb was checked in a clean Debian 13 container. `VERSION` in the tarball says commit f5227b731d2e. - [Draft release](https://gitea.eom.dev/claude/tidewall/releases/tag/v1.0.0) has the release notes, including what doesn't work and that nothing ran on real GPU hardware, plus both `.sha256` files. **Not done: acceptance item 2 (download, verify, run)** The Gitea tools only take uploads as base64 in the call itself, which doesn't work for a 30 MB tarball and a 22 MB .deb. I didn't take an API token from any config to upload some other way. So the binaries aren't attached yet, and I couldn't download them back. **Eric:** the two files are on tower-plus under `build/release-v1.0.0/` in the repo checkout. They're git-ignored, and later `make dist` runs won't overwrite them. - `tidewall-1.0.0-linux-x86_64.tar.gz`, sha256 `5050c0bd…07c26c8` - `tidewall_1.0.0-1_amd64.deb`, sha256 `b9af9e8e…cb614aee` Attach both to the draft and publish it. The next scheduled run will then download them anonymously, check them against the `.sha256` files and run them in a clean `tidewall-` container. Leave it as a draft if you want to wait for your R6 playtest first. Leaving this open with `needs-eric` until the uploaded files are verified. — tidewall-build
claude added the needs-eric label 2026-11-14 23:36:26 +00:00
Author
Owner

Eric, a heads-up before you attach the binaries: the v1.0.0 tag (f5227b7) and the two files in build/release-v1.0.0/ predate #72. They still have the sky barge and wing-sail. Your choice on #72 (the packet ship, lifeline mortar and gull-kite) landed on main in aa79758 today, with every gate passing.

Options:

  1. Recommended: cut v1.0.1 from main. It's a version bump plus make dist-check and make deb-check, and the next run can prepare it as a draft, as before. The v1.0.0 draft could then be deleted. Deleting it is your call, since this task doesn't delete anything on Gitea.
  2. Publish v1.0.0 as it stands and treat the new launch as the first update.

I haven't changed the tag or the draft. Still open with needs-eric.

— tidewall-build

**Eric, a heads-up before you attach the binaries:** the v1.0.0 tag (f5227b7) and the two files in `build/release-v1.0.0/` predate #72. They still have the sky barge and wing-sail. Your choice on #72 (the packet ship, lifeline mortar and gull-kite) landed on `main` in aa79758 today, with every gate passing. Options: 1. **Recommended: cut v1.0.1 from `main`.** It's a version bump plus `make dist-check` and `make deb-check`, and the next run can prepare it as a draft, as before. The v1.0.0 draft could then be deleted. Deleting it is your call, since this task doesn't delete anything on Gitea. 2. Publish v1.0.0 as it stands and treat the new launch as the first update. I haven't changed the tag or the draft. Still open with `needs-eric`. — tidewall-build
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: claude/tidewall#64