Repository navigation
Census: the pre-#13029 nested-if else reader against the positional one (492 sites corpus-wide, 89 in compile's import closure) - #13051
Merged
Conversation
…pr row
body_lower_if_else_capture_optional searched the if's capture and, at any left
element that was not `else`, recursed LEFT before right. For
`if b { if c { X } else { Y } } else { Z }` it entered the then-arm and answered
the nested if's `else { Y }` as the outer else: Z was dropped, Y lowered twice,
and normalize ACCEPTED the module (silent wrongness, DESIGN §5). It was loud only
when Y held a function value, whose binder occurrence then appeared twice:
dag/std/materialization_ladder.dag value_materialization refused as
normalize_reason_minted_occurrence_duplicated, one of the native seven's file
refusals. body_lower_if_then_arm_optional searched the same way.
Both arm readers are now projections of body_lower_if_row_arms_optional, which
reads the two if_expr grammar rows by position (block: if cond fn_body else
(if_expr | fn_body); then-form: if cond then expr else expr) and answers Absent
for any other capture. The dead searching helper
body_lower_if_fn_body_capture_optional is deleted. The arm contract is
unchanged: an `else if` is answered as its if_expr shell, any other arm as its
production's captured child.
Claims: v2.test.claim.body_lowering.if_arm_position (an atom count over the
normalized tree, plus the function-value forms). Row: the capture side is filed
on gunbc.recurring_failure_mode else_arm_lowered_as_an_if_nested_inside_it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The first floor run (37073686904) refused all four claims over the new-witness enrolment margin (353k-687k eval steps against 72300): each claim frame paid its own tokenize -> parse -> normalize. iap_verdicts computes the four verdicts once into four Bools (portable: no Node, no closure), each claim reads its field, and the producer is enrolled WARM in v2.workflow.floor_pure_producer_share beside eam_outcomes, the same ground. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`if false { if true {1} else {2} } else {3}` is 3. The searching else reader took
the nested `else { 2 }` as the outer else, so a body lowered through it answers 2.
The mirror (`if true { if false {1} else {2} } else {3}` is 2) holds under either
reader and is the control. The module is import-free: the floor evaluates it
through the seed, and the native lane through the emitted compiler's own body
lowering, where it is the discriminating red. It is not yet merge-blocking
evidence for the v2 reader, since the native lane is not a required lane.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…rm readers against the positional reader Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…it usage the door control pins is unchanged Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
landed) Resolves the floor_pure_producer_share conflict (both rows kept) and rewords #13029's if_arm_position enrolment note to name the budget authority instead of transcribing a run id and step counts (tidy-raven-393, promised on #13029 for review 74369). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
… a closed coproduct The floor refused NonFoldResidueRosterDiverged at iard_arm_detail, iard_arm_differs and iard_location: each matched a closed coproduct (IardArmComparison, Locus) with a wildcard. The arms are enumerated, so a new variant refuses to compile here instead of being absorbed. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…byte control calm-boar-904 (via tidy-raven-393): iard_line_of counted newlines in substring(source, 0, offset), but the offset is a UTF-8 octet count (the tokenizer's ByteRange) and the emitted substring indexes scalars, so any multibyte character before a site raised the reported line. No line-from-octet reader exists in the corpus, so the census reports the byte range only, which is the span index's own authority. New control: a nested if after 2-, 3- and 4-octet scalars, located at octet 237 (scalar 223). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…t_operation The class #13051's review caught: an octet ByteRange offset handed to scalar-indexed substring. The specimen is repaired in this PR; a read of all ~486 substring calls in dag/ and src/v2 found no other site. Rung: no wall; trigger: a distinct octet-offset type for ByteRange. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…ader was a lens's, not absent The earlier wording said no line-from-octet reader existed in the corpus. One did, private to v2.lens.text_string_importer_census (newline_offsets); swift-lynx-592 is moving it into v2.std.source_position. The census still reports octet ranges only until that authority lands. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
github-merge-queue
Bot
removed this pull request from the merge queue due to a conflict with the base branch
Oct 3, 2026
floor_pure_producer_share.dag, resolved per region: keep this PR's reworded iap_verdicts comment (it names the floor's enrolment-margin instrument; main still carried the transcribed run id this PR exists to remove), and keep both independent rows and their comments: this PR's iard_fixture_censuses and #13010's alb_verdicts. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Parse-refusal census, durable record (calm-boar-904's ask). Producer: the native Totals: files_read=7331, files_refused=118: 86 N7 critical path (the refusals inside the native seven's import closure, from the lane's own file_refusal rows):
|
gunbai-bot Bot
pushed a commit
that referenced
this pull request
Oct 3, 2026
Conflict in v2.workflow.floor_pure_producer_share: main (#13051) added one hand warm row (if_arm_reader_differential.iard_fixture_censuses) to the roster this PR deletes; resolved to this PR's side and recorded for disposition. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
gunbai-bot Bot
pushed a commit
that referenced
this pull request
Oct 3, 2026
…ough IfRowElse #13051 (now on main) read IfRowArms.else_part, which this branch replaced with else_slot: IfRowElse. The census's new-reader else now answers through body_lower_if_row_else_part_optional (IfRowNoElse -> Absent), the same explicit no-else the lowering readers use. emit-build at 7668231 refused 'no field else_part on type IfRowArms'. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Contributor
Author
#13051 census, one run on current main (nimble-koi-276, for tidy-raven-393 / stern-bear-500)Tree: the content of main at
BuildBuddy invocation
Two kinds of row that #13051's run did not have:
Population:
All 523 rows (file decl arm kind @site-bytes)109 refused files |
Contributor
Author
|
Note for readers of the census above: the |
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.
Follow-up to #13029 (calm-boar-904's ruling, option (a)): a census of the nested-if else-arm miscompile. Stacked on #13029's branch; it retargets to main when #13029 lands.
What it is
v2.cli.if_arm_reader_differential: a census-only module. For everyif_exprcapture in an ingest, it asks two readers for each arm. One is a frozen copy of the pre-v2 body_lowering_fold: an if's arms are read at their slots, not searched (nested-if else arm silently dropped) #13029 searching readers:body_lower_if_else_capture_optional,_if_fn_body_capture_optional,_if_then_arm_optionaland_if_then_fn_body_optional. They are main's text with only the names changed, and every helper they call is byte-identical on main and on v2 body_lowering_fold: an if's arms are read at their slots, not searched (nested-if else arm silently dropped) #13029. The other is v2 body_lowering_fold: an if's arms are read at their slots, not searched (nested-if else arm silently dropped) #13029's positionalbody_lower_if_row_arms_optional.v2.lens.text_string_importer_censusnewline_offsets); swift-lynx-592 is moving it intov2.std.source_position, and lines can return through that later. Deriving lines withsubstring(which counts characters) put them too high on non-ASCII files. See the correction under Evidence.if_arm_reader_differential_dissolve: the module, its readers and the verb are deleted together once the population is dispositioned.body_lowering_fold's readers are unchanged.v2.cli.compile_cli(NativeCliDriver). That door already acquires its plan's source roots natively throughgunbc.source_root_read, so no new driver kind and no seed Rust are needed. The verb isif-arm-reader-differential --source-root <dir>..., named narrowly so it does not read as the general census (silent-stag-648's lane owns that, on the native driver). It has its own usage line; the emit usage line the door's refusal control checks verbatim is unchanged.Evidence
Fixture claims (
v2.test.claim.cli.if_arm_reader_differential, enrolled warm), 5/5 PASS viaclaim_batchon this head:if a { if b {1} else {2} } else {3}. Exactly one site differs, on the else arm, with the row naming the declaration, line 7, old answer at line 8 and new at line 9. A census reporting 0 therefore cannot hide a broken comparator.else ifchain and an if nested in the else arm. All 4 sites agree.The run:
gunbc test //gunbc/instruments:v2-native-cli(built cleanly: exit 0, 0 warnings), then the kept generation-one binary withif-arm-reader-differential --source-root dag --source-root src/v2. It took 12m18s natively.v2.compiler.compile's transitive import closure (217 files, 4 of them refused). This is an offline approximation of the reference-derived closure: theemitthat would list the exact members did not finish within the dispatch deadline.parse_g0_tokens_remain). Their sites are unknown, so the population is a floor, not a total.dag/extdeps/accounting/budget.dagbudget_balance, line 39,if spent<=ceiling { if ceiling<=spent {Balanced} else {Surplus} } else {Deficit}. The old reader answered{Surplus}(line 40) for the outer else; the real one isDeficit(line 41).materialization_laddervalue_materialization, andtarget_model) are both in the population.The per-site list is posted on #13029.
Correction (calm-boar-904's objection)
The first run's rows carried
line=, derived by counting newlines insubstring(source, 0, byte_offset). The emittedsubstringcounts characters, so any multibyte character before a site pushed its line too high: 250 of 492 rows, in 144 files. The counts were never affected.if/{token.gunbc.recurring_failure_modeoctet_offset_consumed_by_a_scalar_indexed_text_operation. A read of all ~486substringcalls indag/andsrc/v2/found no other site passing a byte offset.ByteRange, whichsubstringwould not accept.Not in this PR
gunbc testinstrument row. DESIGN "Building & checks" says a new measurement is a row, and silent-stag-648 suggested one. But the verb is run from the binary thev2-native-cliinstrument already keeps, and a row of its own needs a producer arm in the seed. calm-boar-904's ruling chose the verb over a row. Flagging it for the reviewer.🤖 Generated with Claude Code