Skip to content

Keep a literal binding distinct from a dummy with the same bits - #178

Open
dcharlot65-openie wants to merge 1 commit into
cursor/mset-flag-index-1893from
cursor/maps-empty-text-1893
Open

dcharlot65-openie wants to merge 1 commit into
cursor/mset-flag-index-1893from
cursor/maps-empty-text-1893

Conversation

@dcharlot65-openie

@dcharlot65-openie dcharlot65-openie commented Nov 13, 2026 •

Copy link
Copy Markdown
Contributor

A var or let of a literal reused a capture when the bits matched. The decimal helpers in fn freq bind empty text that way. The branch that returns the text also has a dummy initial of empty text. The schedule read the binding back as that parent dummy, so construct(schedule(F)) was not F. The gate was not the cause: both forms constructed.

A binding is now its own node. A literal already built in the same region is still that node. An operand still shares a capture, so a domain test and the operation it guards stay one value. FG7, FG8, and C7 are unchanged.

fn freq's closure round-trips. 21_maps.rce is option B. The whole module's first refusal is still exit.

The same distinction clears other closures whose mismatch was a binding and a dummy of the same bits. The syntactic floor stays 5. Four more modules schedule back wholly: 08_traits.rce, 18_while_bound.rce, 24_lambda.rce, 27_bitops.rce.

Census

0208c624 head ec6a6d23
programs 60 60
floor 5 5
option B 0 12
whole 2 6
whole-entry-only 1 1
whole-with-user 1 5
first-tag break 18 18
first-tag catch 0 0
first-tag FG8 none none
first-tag defect 3 4 (22_tuples.rce is now the file's first tag)
first-tag round-trip 19 6
21_maps.rce whole exit; roots round-trip whole exit; roots clear (option B)

First-tag round-trip on 0208c624: 03_fib_loop.rce, 08_traits.rce, 10_comptime.rce, 11_arena.rce, 18_while_bound.rce, 19_floats.rce, 21_maps.rce, 22_tuples.rce, 24_lambda.rce, 25_labels.rce, 27_bitops.rce, 28_slices.rce, 33_in_operator.rce, 36_multiline_brackets.rce, 41_arena_bound_wrapping.rce, 43_simd_lanes.rce, 45_byte_text.rce, 46_u64.rce, 49_loop_limit.rce.

First-tag round-trip on ec6a6d23: 10_comptime.rce, 11_arena.rce, 19_floats.rce, 28_slices.rce, 33_in_operator.rce, 41_arena_bound_wrapping.rce.

Of the nine names that were in that list on both earlier revisions, 03_fib_loop.rce, 25_labels.rce, 36_multiline_brackets.rce, 43_simd_lanes.rce, 45_byte_text.rce, 46_u64.rce, and 49_loop_limit.rce are now option B. 10_comptime.rce stays round-trip. 22_tuples.rce's first tag is defect because an older refusal is now the first one. The defect itself is unchanged. See below.

22_tuples.rce is case (b)

The file has two program roots. fn divmod sorts first. The file's first tag is that root's refusal, or the next root's if fn divmod applies.

On 0208c624, fn divmod constructs and its schedule constructs. The bytes differ, so the tag is round-trip. There is no defect on that root. fn use_it is the next root. Its schedule refuses, on that same revision:

requires change schedule -: defect: the schedule is not well formed: in ac28239692a3f0a647c19ea503e3d46a640e6487f3ae6d125209e2ee06b0beb3/0: W3: the paths that meet after an if hold different owned values ({1: Live, 3: Dead, 4: Live} and {1: Dead, 3: Dead, 4: Live})

On ec6a6d23, fn divmod round-trips (construct(schedule(F)) is F). fn use_it refuses with that same sentence, including the same group key and the same live and dead sets. The defect was already reachable by scheduling fn use_it directly. It was not the file's first tag while fn divmod still mismatched.

Not merged. Not deployed. PR 172's branch is untouched.

Open in Web Open in Cursor 

A var or let of a literal reused a capture of those bits. The decimal
helpers' empty text is that binding, and a branch output's dummy is the
same bytes, so the schedule read the text back in the parent. The binding
is now its own node. An operand still shares a capture, so a domain test
and the operation it guards stay one value.

fn freq's closure round-trips. 21_maps.rce is option B. Its whole module
is still exit.

Co-authored-by: David Charlot <dcharlot65-openie@users.noreply.github.com>

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants