fix(build): //base belongs in public_deps where it is in the public interface #57

Merged

triform-admin merged 1 commit from fix/base-public-deps into main

2026-07-30 21:21:23 +00:00

triform-admin commented

2026-07-30 17:37:15 +00:00

Owner

A build-graph defect that only a unit test could see — found by the build I fired to prove #56's tests pass.

What happened

That build didn't get to running tests. It failed to compile one:

FAILED: cb_wire_envelope_test.o
../../base/memory/raw_ptr_exclusion.h:11:10: fatal error:
  'partition_alloc/pointers/raw_ptr_exclusion.h' file not found

:cb_wire_envelope listed //base in private deps. But cb_wire_envelope.h includes base/values.h in its public interface (ProbeResultPayload holds a base::DictValue), so anything including that header needs //base's include dirs transitively — and //base is what pulls in partition_alloc's generated headers.

The source_set itself compiled fine; only dependents broke. With no dependent building, nothing could observe it. That's gn's deps vs public_deps distinction exactly: a dep whose headers appear in your own public headers is part of your interface.

It was a class, not an instance

I scanned every source_set in capture/ for "public header includes base/ AND //base is a private dep". Nineteen targets, including:

pointer_state, input_dispatch_ime, input_dispatch_touch, input_dispatch_drag, input_dispatch_clipboard, cursor_client, cursor_xy_join, embedder, framesink_capture, stats_relay, clipboard_relay, file_upload_relay, …

Note pointer_state and input_dispatch. Their test binaries — cloud_browser_pointer_state_unittests, cloud_browser_input_dispatch_unittests — are two of the five that have never been built in any lane. This is very likely why: whoever last tried hit the same wall and dropped the target rather than the dep. I can't prove that from here (no record of the attempt), but the shape matches.

Nothing was broken in production: the worker links all of these transitively either way. The defect is only visible to a target that depends on one of them in isolation — i.e. a unit test with narrow deps. That's why it survived while the tests stayed unbuilt.

Verified

t7 build on this branch, with cb_wire_envelope_unittests added to the targets:

[19/654] 10.93s F CXX obj/.../cb_wire_envelope_unittests/cb_wire_envelope_test.o
[31/651] 12.80s F LINK ./cb_wire_envelope_unittests
12m10.27s Build Succeeded: 620 steps

It compiles and links where it previously died.

One caveat I want to be explicit about: STEP 7 in that build still only invoked the two hardcoded binaries, so the wire-envelope assertions have still not executed. That's not this PR's job — #56 is what teaches STEP 7 to derive its list. The two changes are complementary and both are needed: this one makes the test compile, #56 makes it run. Neither alone closes the gap.

A build-graph defect that only a unit test could see — found by the build I fired to prove #56's tests pass. ## What happened That build didn't get to running tests. It failed to **compile** one: ``` FAILED: cb_wire_envelope_test.o ../../base/memory/raw_ptr_exclusion.h:11:10: fatal error: 'partition_alloc/pointers/raw_ptr_exclusion.h' file not found ``` `:cb_wire_envelope` listed `//base` in private `deps`. But `cb_wire_envelope.h` includes `base/values.h` in its **public** interface (`ProbeResultPayload` holds a `base::DictValue`), so anything including that header needs `//base`'s include dirs transitively — and `//base` is what pulls in partition_alloc's generated headers. The source_set itself compiled fine; only **dependents** broke. With no dependent building, nothing could observe it. That's gn's `deps` vs `public_deps` distinction exactly: a dep whose headers appear in your own public headers is part of your interface. ## It was a class, not an instance I scanned every `source_set` in `capture/` for *"public header includes `base/` AND `//base` is a private dep"*. **Nineteen targets**, including: > `pointer_state`, `input_dispatch_ime`, `input_dispatch_touch`, `input_dispatch_drag`, `input_dispatch_clipboard`, `cursor_client`, `cursor_xy_join`, `embedder`, `framesink_capture`, `stats_relay`, `clipboard_relay`, `file_upload_relay`, … Note `pointer_state` and `input_dispatch`. Their test binaries — `cloud_browser_pointer_state_unittests`, `cloud_browser_input_dispatch_unittests` — are **two of the five that have never been built in any lane**. This is very likely why: whoever last tried hit the same wall and dropped the target rather than the dep. I can't prove that from here (no record of the attempt), but the shape matches. Nothing was broken in production: the worker links all of these transitively either way. The defect is only visible to a target that depends on **one** of them in isolation — i.e. a unit test with narrow deps. That's why it survived while the tests stayed unbuilt. ## Verified t7 build on this branch, with `cb_wire_envelope_unittests` added to the targets: ``` [19/654] 10.93s F CXX obj/.../cb_wire_envelope_unittests/cb_wire_envelope_test.o [31/651] 12.80s F LINK ./cb_wire_envelope_unittests 12m10.27s Build Succeeded: 620 steps ``` It **compiles and links** where it previously died. One caveat I want to be explicit about: STEP 7 in that build still only *invoked* the two hardcoded binaries, so the wire-envelope assertions have **still not executed**. That's not this PR's job — #56 is what teaches STEP 7 to derive its list. **The two changes are complementary and both are needed**: this one makes the test compile, #56 makes it run. Neither alone closes the gap.

triform-admin added 1 commit

2026-07-30 17:37:15 +00:00

fix(build): //base belongs in public_deps where it is in the public interface

CodeQL / Analyze go (pull_request) Has been skipped

Details

CI / Docs link check (pull_request) Successful in 15s

Details

CodeQL / Analyze javascript-typescript (pull_request) Has been skipped

Details

native-peer-gate / native-peer-gate-scaffold (permissive) (pull_request) Successful in 39s

Details

native-peer-gate / native-peer-gate-strict (M7 gate) (pull_request) Successful in 43s

Details

E2E / docker-compose + Playwright (pull_request) Successful in 48s

Details

CI / Container smoke test (pull_request) Successful in 7m10s

Details

CI / Lint (pull_request) Successful in 15m42s

Details

5811ab8d7a

The build I fired to prove cb_wire_envelope_unittests passes did not get that
far — it failed to COMPILE the test, with:

  FAILED: cb_wire_envelope_test.o
  ../../base/memory/raw_ptr_exclusion.h:11:10: fatal error:
    'partition_alloc/pointers/raw_ptr_exclusion.h' file not found

Root cause: :cb_wire_envelope listed //base in private `deps`. But
cb_wire_envelope.h includes "base/values.h" in its PUBLIC interface
(ProbeResultPayload holds a base::DictValue), so any target that includes
that header needs //base's include dirs transitively — and //base is what
pulls in partition_alloc's generated headers. The source_set itself compiled
fine (it sees //base directly); only DEPENDENTS broke. With no dependent
building, nothing could observe it.

That is gn's deps-vs-public_deps distinction exactly: a dep whose headers
appear in your own public headers is part of your interface, not an
implementation detail.

Then the obvious question — is this one target or a class? Scanned every
source_set in capture/ for "public header includes base/ AND //base is a
private dep". NINETEEN targets, including:

  pointer_state, input_dispatch_ime, input_dispatch_touch,
  input_dispatch_drag, input_dispatch_clipboard, cursor_client,
  cursor_xy_join, embedder, framesink_capture, stats_relay,
  clipboard_relay, file_upload_relay, active_webcontents_resolver,
  cb_audio_lifecycle, cb_signaling_ws_client, cb_offerer_driver,
  cb_dc_host, cb_signaling_reconnect, cb_wire_envelope

Note which names appear there: pointer_state and input_dispatch. Their test
binaries — cloud_browser_pointer_state_unittests,
cloud_browser_input_dispatch_unittests — are two of the five that have never
been built in any lane. This is very likely why: whoever last tried to add
them hit this same wall and dropped the target rather than the dep. I cannot
prove that from here (no record of the attempt), but the shape matches.

All nineteen moved to public_deps. The worker binary links them all
transitively either way, which is why nothing was broken in production — the
defect is only visible to a target that depends on ONE of them in isolation,
i.e. a unit test. That is the whole point of unit tests having narrow deps,
and it is why this stayed hidden while the tests stayed unbuilt.

Verified locally: brace/bracket balance on both BUILD.gn files, no target
left with //base in both deps and public_deps, re-scan reports zero remaining
affected targets, make verify + cxx-include-lint clean.
NOT verified: that it compiles. A t7 build follows — that is the point.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>