[CI][Bench] Publish to the snapshots branch, not over the repository - #2009
Merged
Conversation
The publish replaces everything in the branch it writes, which on `main` meant the first run deleted TileOPs-nightly's README and licence. Those belong to whoever maintains that repository, not to this job. The snapshot goes to a `snapshots` branch of its own: one commit per run, three files, and everything not in the payload deleted — which is what keeps a stale test_results.xml from passing for tonight's. `main` is left to hold what the repository says about itself.
lcy-seso
force-pushed
the
bench/nightly-data-branch
branch
from
August 30, 2026 04:17
b8f5ae7 to
120f70a
Compare
lcy-seso
added a commit
to tile-ai/TileOPs.github.io
that referenced
this pull request
Aug 30, 2026
…ents (#40) Pairs with tile-ai/TileOPs#2009. Either order is safe: the `snapshots` branch already holds the snapshot from run 33289823876. ## Read from the snapshots branch The nightly publishes by replacing everything in the branch it writes. Pointed at `main`, the first run deleted TileOPs-nightly's README and licence — those belong to whoever maintains that repository. It now writes a `snapshots` branch, and this reads from there: same three files, one commit per run. ## An op's heading is its name The heading carried `(3 workloads · ✅)`, and the heading is what the page's table of contents shows — 189 entries, 88 of them ending in a tick that says the op passed, which is what a reader assumes anyway. The workload count is the length of the list immediately under it. What survives is the mark that warns: ❌ where the correctness run failed, ⏭️ where it was skipped. On the last snapshot that is no op at all. The golden pages carry the change, and the fixture covers all three cases. ## Say each design decision once The comments here argue with themselves: nearly every one states a decision, restates it as the case against the alternative, then names the symptom that alternative produced. Compressed, not deleted — every comment kept was tested against one question: would removing it let someone undo the decision by accident? `workload_shape`'s module docstring also stops describing a merge that does not happen: templates are read only where no shape is given outright, all or nothing, which is what the code has always done. 847 comment lines to 722; the stylesheet from 421 to 340. ## Checks `pytest`, `ruff`, `stylelint`, `mkdocs build` pass. Rendered against the live branch: 189 ops, 1139 workloads.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follows #2008. Pairs with tile-ai/TileOPs.github.io#40; either order is safe.
Problem
The publish replaces everything in the branch it writes. Pointed at
main, the first run deleted TileOPs-nightly's README and licence — those belong to whoever maintains that repository, not to this job.Change
The snapshot goes to a
snapshotsbranch of its own: one commit per run, three files, and everything not in the payload deleted there — which is what keeps a staletest_results.xmlfrom passing for tonight's.mainholds what the repository says about itself, and nothing publishes over it.The branch already carries the snapshot from run 33289823876, and TileOPs-nightly's
mainhas been restored.