Record CR-2026-034 in SESSION-NOTES.md - #35
Merged
Conversation
The RESUME HERE block named `flagged` in the C interpreter as the one obvious next CR. It landed, so that entry is replaced with what the harness now says the next work is: widening the harness rather than the interpreter. `repeat` is the last construct C has no field type for and it is 3 schemas, worth less than the two harness limits above it - `transform` at 26 and `bitfield_string` at 24 - and whether C supports either is unknown *because the harness cannot build them*, which is the question the harness exists to answer. The C bullet now lists what blocks each remaining schema **with the side the limit sits on**, because CR-2026-034 corrected a message that did not: `no constructor for type 'u32le16'` read as a C gap when C decodes it perfectly well. In a report whose purpose is telling C's gaps from the harness's, a reason that does not say which is worse than no reason. The session subsection gains CR-2026-034: the representation reuses the tlv shape, and the one thing that shape did not cover - a `flagged` mask field must declare `var_name`, because the interpreter records a variable only where a field declares one while a YAML `flagged` names a field. `var_has()` makes a missing reference an error rather than a mask of zero, which would have decoded nothing and reported success. The measurements section gains a ninth kind, and it is a new failure mode rather than a repeat: a multi-edit script that asserts anchors as it goes and writes at the end leaves the file untouched when a later anchor fails, and the `--amend` that follows commits the unchanged content. The edit silently had not happened and the commit looked fine. It caught me twice in a row on AGENTS.md in the same CR. Counts: Python 2706, harness 488 of 488. Co-Authored-By: Claude Opus 5 (1M context) <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
1 participant
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.
Notes-only. No code, schema or test changed -
git diff --stat masteris one file.What was stale
The
RESUME HEREblock namedflaggedin the C interpreter as "one obvious next CR". It landed as CR-2026-034, so the entry is replaced with what the harness now says the next work actually is:repeatis the last construct C has no field type for, and it is 3 schemas - worth less than the two limits above it,transform(26 schemas) andbitfield_string(24). Whether C supports either is unknown because the harness cannot build them, which is the question the harness exists to answer.Counts corrected: Python 2706 passed / 4 skipped,
make test-c488 of 488 attempted, 0 differ.The C bullet now names the side each limit sits on
CR-2026-034 corrected a skip message that did not:
no constructor for type 'u32le16'read as a C gap when the interpreter hasFIELD_TYPE_U32LE16and decodes it perfectly well. In a report whose whole purpose is telling C's gaps from the harness's, a reason that does not say which is worse than no reason at all.Also records the rule that shape did not cover: a
flaggedmask field must declarevar_name. The interpreter records a variable only where a field declares one, while a YAMLflaggednames a field - so no corpus schema carriesvar:for it and the harness patches it in.var_has()makes a missing referenceSCHEMA_ERR_MATCHrather than a mask of zero, which is indistinguishable from "no bits set" and would have decoded nothing while reporting success.A ninth failure kind
New mode, not a repeat of the eight: a multi-edit script that asserts anchors as it goes and writes at the end leaves the file untouched when a later anchor fails, and the
git commit --amendthat follows commits the unchanged content. The edit silently had not happened and the commit looked clean. It caught me twice in a row on AGENTS.md within one CR. Verify every anchor before writing any.The section opener no longer carries a running instance total - it went stale within the same day, which is the very trap the section is about.
Verified
python3 -m pytest tests/ -q-> 2706 passed, 4 skippedmake docs-index-check-> up to datetests/,tools/, theMakefileor any language runner readsSESSION-NOTES.md(the one grep hit ingenerate_output_schema.pyis a prose comment)🤖 Generated with Claude Code