feat(bazel): SOCKET_BAZEL_FORCE_QUERY_FALLBACK env-var gate for deterministic fallback-parser coverage

[Simon (simonhj)]

Summary

Add a SOCKET_BAZEL_FORCE_QUERY_FALLBACK env variable that, when truthy (1 / true / yes, case-insensitive), makes socket manifest bazel skip the unsorted_deps.json fast path and parse extraction output through the bazel query --output=build regex fallback instead.

Why

The Bazel extractor has two parser paths inside extractFromOneRepo:

1. Fast path — JSON.parse over <externalDir>/<repo>/unsorted_deps.json (~50 lines)
2. Fallback path — six regex patterns over the cached bazel query --output=build stdout, in bazel-build-parser.mts

Today the choice is decided purely by existsSync(unsorted_deps.json). Whichever path Bazel happens to materialize on disk runs. The fast path is also the path that runs in 90%+ of real corpus repos, because rules_jvm_external materializes unsorted_deps.json by default.

What changed

• New isForceQueryFallbackEnabled() helper in extract_bazel_to_maven.mts — reads process.env['SOCKET_BAZEL_FORCE_QUERY_FALLBACK'], normalizes case, treats 1 / true / yes as on.
• extractFromOneRepo now early-returns from the unsorted_deps.json candidates loop when the gate fires, falling through to parseBazelBuildOutput(cachedProbeStdout).
• A logger.verbose line is emitted when the gate fires (only when queryOpts.verbose is set).
• Eight new test cases in extract_bazel_to_maven.test.mts covering:
  • Env unset → fast path used (existing behavior preserved)
  • 1 / true / YES (case-insensitive) → fallback used
  • '' / 0 / false → fast path used (falsy)
  • Two parsers pitted against each other (each produces a coordinate the other doesn't) to assert which one ran by inspecting the manifest output.

---

Note

Low Risk
Default behavior is unchanged unless SOCKET_BAZEL_FORCE_QUERY_FALLBACK is set; the main risk is altering artifact extraction results when the env var is enabled and bypasses unsorted_deps.json.

Overview
Adds SOCKET_BAZEL_FORCE_QUERY_FALLBACK (truthy: 1/true/yes) to make extractFromOneRepo skip the unsorted_deps.json fast path and always fall through to parsing cached probe stdout via the regex fallback, with a verbose diagnostic when enabled.

Extends unit tests to assert fast-path vs fallback selection across truthy/falsy env values by seeding both sources with distinct coordinates and verifying which lands in the generated maven_install.json.

^{Reviewed by Cursor Bugbot for commit a8d0322. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Tests read manifest from non-existent _whole_repo subdirectory
  • Removed the incorrect '_whole_repo' path segment from all four readFileSync calls in the SOCKET_BAZEL_FORCE_QUERY_FALLBACK tests so they read from the correct output location (path.join(tmp, 'maven_install.json')).

Or push these changes by commenting:

@cursor push f15aaf1792

Preview (f15aaf1792)

diff --git a/src/commands/manifest/bazel/extract_bazel_to_maven.test.mts b/src/commands/manifest/bazel/extract_bazel_to_maven.test.mts
--- a/src/commands/manifest/bazel/extract_bazel_to_maven.test.mts
+++ b/src/commands/manifest/bazel/extract_bazel_to_maven.test.mts
@@ -517,7 +517,7 @@
 
     expect(result.ok).toBe(true)
     const manifest = JSON.parse(
-      readFileSync(path.join(tmp, '_whole_repo', 'maven_install.json'), 'utf8'),
+      readFileSync(path.join(tmp, 'maven_install.json'), 'utf8'),
     )
     // The JSON parser ran: from-json coord is present, from-regex is absent.
     expect(manifest.artifacts['com.example:from-json']).toBeDefined()
@@ -539,7 +539,7 @@
 
     expect(result.ok).toBe(true)
     const manifest = JSON.parse(
-      readFileSync(path.join(tmp, '_whole_repo', 'maven_install.json'), 'utf8'),
+      readFileSync(path.join(tmp, 'maven_install.json'), 'utf8'),
     )
     // The regex parser ran: from-regex coord is present, from-json is absent.
     expect(manifest.artifacts['com.example:from-regex']).toBeDefined()
@@ -573,7 +573,7 @@
       expect(result.ok).toBe(true)
       const manifest = JSON.parse(
         readFileSync(
-          path.join(tmp, '_whole_repo', 'maven_install.json'),
+          path.join(tmp, 'maven_install.json'),
           'utf8',
         ),
       )
@@ -600,7 +600,7 @@
       expect(result.ok).toBe(true)
       const manifest = JSON.parse(
         readFileSync(
-          path.join(tmp, '_whole_repo', 'maven_install.json'),
+          path.join(tmp, 'maven_install.json'),
           'utf8',
         ),
       )

_{You can send follow-ups to the cloud agent here.}

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit a8d0322. Configure here.}