Repository navigation
Reuse cached scores on nodes with another st_dev - #35
Merged
Merged
Conversation
file_fingerprint() put the input files' device number into the stat-mode cache identity. A shared filesystem (BeeGFS on the EMBL cluster) reports a different st_dev for the same file on each node, so a recursive run on one node treated every interfaces.csv certified on another as stale and recomputed it, although size, mtime, ctime and inode were identical. In a 1,900-fold AlphaPulldown campaign, 890 of 1,913 manifests differed only in st_dev (65 vs 59), and the summary report spent days rescoring valid tables. Size, mtime_ns, ctime_ns and inode stay in the fingerprint, so edits and replaced files are still detected (the replaced-inputs test is unchanged). Manifests written by 1.4.4 are recomputed once, as after any software change, because the software identity hashes the package source. The new regression test scores a run, then reports another st_dev for every file and checks that the table is reused; it fails on main. Full suite: 183 passed, 6 skipped. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PAE plot fingerprints use the same file_fingerprint(), so a recursive run on another node also redrew every plot it found: the AlphaPulldown campaign's summary report redrew ~10,000 plots before the device number was dropped. The regression test now draws plots on the first run and fails if the second run, with another st_dev, reloads a model to redraw one; it fails on main. The CHANGELOG entry mentions plots. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DimaMolod
marked this pull request as ready for review
November 2, 2026 08:31
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
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.
Stat-mode cache validation included
st_devin each file fingerprint. On a shared filesystem the same file can have a differentst_devon each node (BeeGFS on the EMBL cluster: 65 on one node, 59 on another). Soalphajudge --recursiveon one node recomputed every table certified on another, and redrew every PAE plot, although size, mtime, ctime and inode matched.Seen in a 1,900-fold AlphaPulldown run:
st_dev;Change
file_fingerprint()no longer recordsdevice. It covers both inputs and PAE plots. Size, mtime_ns, ctime_ns and inode remain, so edits and replaced files are still detected.st_devfor every file, and check that neither the table nor any plot is redone. It fails onmain.Manifests written by 1.4.4 are recomputed once after upgrading, as after any software change.
Tests: full suite 183 passed, 6 skipped. After the plot test was added,
test_run_integrity.pypasses in full (59).🤖 Generated with Claude Code