upb's reflection sources include only their internal header:
upb/reflection/field_def.c includes upb/reflection/internal/field_def.h,
which reaches the upb/reflection/field_def.h it implements through def.h.
So the translation-unit rule of PR 2684 never read field_def.c as that
header's implementation, and 343 calls from protobuf's Python extension,
plus 88 from its Lua binding, still went to the copy of upb that the PHP
extension bundles in php-upb.c.
When a source includes no header named like it in its own folder, one
there that it includes through other headers counts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PR #2675 narrowed a C or C++ test's same-named candidates to the ones its
translation unit declares. Production code now gets the same rule:
protobuf's Python extension called the copies of upb's C API that the PHP
extension bundles in php-upb.c, and its Java generator's calls of the free
HasHasbit() went to ImmutableFieldGenerator::HasHasbit, a method beside the
caller.
Refinements the wider scope needed:
- a source implements the same-stem header beside it, when it includes one:
pyext/message.cc is no implementation of google/protobuf/message.h;
- a header's own overloads reach the definitions its source holds:
conformance/binary_wireformat.h's template Varint calls the .cc's;
- narrowing never leaves a call with only functions its arguments can't
fill, past one outside the unit that they can: glog's three-argument
GetStackTrace(...) is not utilities.h's GetStackTrace(). This also mends
9 breakpad unittest calls #2675 had narrowed the same way;
- nor, for a C call, to definitions only a flag nothing sets compiles,
past one the build without it does: that stays #2615's choice.
A file-level "call" (a macro, or a declaration read as a call) is left as
it was.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
