Repository navigation
test(fork): run the readerAfter pinned-WAL test with 1 KiB pages - #206
Merged
fireboltdude1357 merged 1 commit intoNov 6, 2026
Merged
fireboltdude1357 merged 1 commit into
fireboltdude1357 merged 1 commit into
Conversation
At upstream's 32,768-frame ceiling and 4 KiB pages, the WAL reaches 128 MiB before the ceiling. On omarchy, the readerAfter case's crossing write was then reported blocked twice while the worker checkpointed it, which the test reads as a held write. With 1 KiB pages the ceiling sits near 32 MiB, as with the old 10,000-frame ceiling, and the row counts go back near their old values. The TRUNCATE retry test, which passed on omarchy, keeps 4 KiB pages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The fork's nightly build still fails on the Linux leg after #205:
Write 127 is the one that crosses upstream's 32,768-frame ceiling, and in this case the reader opens only after it. So this isn't the pinned-reader case. With 4 KiB pages the WAL is 128 MiB at the ceiling. On omarchy the checkpoint worker took long enough to report that write blocked twice, which the test treats as a held write. With the old 10,000-frame ceiling the same crossing happened at about 40 MiB, and omarchy passed.
The
readerAftertest now setsPRAGMA page_size = 1024before WAL mode. A 1 MiB row is then about 1,030 frames, so the 32nd row crosses the ceiling, with about 32 MiB of WAL. The reader opens after 31 rows, just past the ceiling, asreaderAfter: 38did before. The loop is back to 64 rows. The TRUNCATE retry test passed on omarchy, so it keeps 4 KiB pages and 126 rows. With 1 KiB pages it timed out here.On a stack of upstream 98b930f with all 23 patches and this branch, the 26 tests in
nodeSqliteClient.test.tspass three times in a row on stl-wsl, in about 16 s. The fork build runs them on omarchy.Made by Claude Opus 5.5 in Claude Code.
🤖 Generated with Claude Code