docs(clickhouse): the value-position NOT (NOT bug has a fix in flight - #14
Open
alexey-milovidov wants to merge 1 commit into
Open
docs(clickhouse): the value-position NOT (NOT bug has a fix in flight#14alexey-milovidov wants to merge 1 commit into
alexey-milovidov wants to merge 1 commit into
Conversation
Copy link
Copy Markdown
Member
Sorry, something went wrong.
This entry's diagnosis was right down to the line: the `name == "not"` branch of `cloneDAGWithInversionPushDown` pushes an inversion through every `not` regardless of `boolean_context`, so a pair of them cancels and `NOT NOT c1` reaches index analysis as bare `c1`. ClickHouse#116381 gates that branch on `boolean_context`, so `not` stays an ordinary function in a value position - the atom is then opaque to index analysis, which is sound - while a truth-tested `NOT` is still pushed down as before. Record the PR so the next triage does not re-derive it, and note what to expect while it is unmerged: it accounted for 20 of the 87 findings in the 2026-08-25 ClickHouse nightly, spread over eight oracles here plus SQLancer++'s NoREC and QUERY_PARTITIONING, so any of those with `NOT (NOT` in the failing query is this one bug rather than ten. Per this file's own convention the entry gets deleted once the fix merges and head stops reproducing.
Copy link
Copy Markdown
Member
Author
|
@qoega could you review this one? A formal review request is not possible here - GitHub only allows them from collaborators on this repository, and you are not one on ClickHouse/sqlancer yet. |
Sorry, something went wrong.
qoega
approved these changes
Aug 26, 2026
qoega
left a comment
Copy link
Copy Markdown
Member
There was a problem hiding this comment.
why do we have claude md in this repo? and not in CI check folder?
Sorry, something went wrong.
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
Reviewers
1 more reviewerAt least 1 approving review is required to merge this pull request.
2 participants
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 NOT (NOT key) entry in the known-open-bugs list was right down to the line — the
name == "not" branch of cloneDAGWithInversionPushDown pushes an inversion through every not
regardless of boolean_context, so a pair of them cancels and NOT NOT c1 reaches index analysis
as bare c1.
ClickHouse#116381 gates that branch on
boolean_context: not stays an ordinary function in a value position, where the atom is then
opaque to index analysis (sound), while a truth-tested NOT is still pushed down as before. It
carries a regression test covering the statistics, view, primary-key, partition and minmax
skip-index pruning paths, and I diffed all 233 index/pruning-related stateless tests with and
without the change — the new test is the only one whose result moves.
So this updates the entry rather than deleting it (the fix is not merged yet), and records what to
expect until it is: the bug accounted for 20 of the 87 findings in the 2026-08-25 ClickHouse
nightly, spread over KeyCondition, ViewEquivalence, SubqueryMaterialize, PartitionMirror,
CountOptimization, TLPGroupBy, TLPDistinct, TLPAggregate here plus SQLancer++'s NoREC and
QUERY_PARTITIONING. Any of those with NOT (NOT in the failing query is this one bug, not ten —
which is the thing that cost the most time to work out during triage.
Per this file's own convention the entry gets deleted once the fix merges and head stops
reproducing.