Skip to content

Take vsa from the repository it was migrated to - #686

Merged
gHashTag merged 2 commits into
mainfrom
feat/depend-on-zig-hdc
Aug 11, 2026
Merged

Take vsa from the repository it was migrated to#686
gHashTag merged 2 commits into
mainfrom
feat/depend-on-zig-hdc

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

src/vsa.zig moved to gHashTag/zig-hdc in 42490a2 and the two references here were left pointing at empty space — one of the reasons this build has failed since March.

It could not be repointed at the time because that package could not be depended on by anything: manifest opened with {, bare identifier where an enum literal is required, no .fingerprint, dependency url with no .hash, and addTest passed root_source_file (removed in 0.15). Repaired in gHashTag/zig-hdc#1, which now has CI on the version this repo targets.

Two lines, not eleven

None of the eleven files in src/ that import vsa.zig is reachable from any of the 133 build roots under src/. Their broken imports have never mattered. Only build.zig named the file — a test target and a module for TRI. Measured before editing; the alternative plan was rewriting eleven consumers that nothing compiles.

The ratchet earned its keep

First run after the edit:

FAIL 1 path(s) in the baseline now exist and must be removed from it:
  src/vsa.zig

Baseline drops 5 → 4.

src/vsa.zig moved to gHashTag/zig-hdc in 42490a2 and the two references here
were left pointing at the empty space -- one of the reasons this build has
failed since March. The module is intact there and exports all fourteen symbols
this repository asked of the file it lost.

It could not simply be repointed at the time, because that package could not be
depended on by anything: its manifest opened with { instead of .{, named itself
with a bare identifier where an enum literal is required, carried no
fingerprint, and gave its own dependency a url with no hash -- and its build
script passed root_source_file to addTest, which stopped being a field in 0.15.
Repaired in zig-hdc#1, which now has CI on the version this repository targets.

The change here is two lines, not eleven, and the reason is worth recording:
none of the eleven files in src/ that import vsa.zig is reachable from any of
the 133 build roots under src/. Their broken imports have never mattered. Only
build.zig itself named the file -- a test target and a module for TRI -- so
only those two needed repointing. Measured before editing rather than assumed;
the alternative plan was to rewrite eleven consumers that nothing compiles.

The baseline drops from five to four, and the ratchet demanded it: a path that
has been repaired and left in the file fails the check, which is what happened
on the first run after this edit.
addTest requires its root module to carry a known target, and a module exported
by a dependency does not -- it takes its target from whatever compiles it, so
passing it straight to addTest panics in the build runner.

Rebuilding one locally from the dependency's source would mean re-declaring
that package's own imports here and testing somebody else's code from the
outside. gHashTag/zig-hdc runs zig build test over exactly this module in its
own CI, on 0.15.2, the version this repository targets -- so the tests execute
where the code lives and against the compiler that will consume it.
@gHashTag
gHashTag merged commit 486de29 into main Aug 11, 2026
15 of 27 checks passed
@gHashTag
gHashTag deleted the feat/depend-on-zig-hdc branch August 11, 2026 01:34
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.

1 participant