I realized that headerscheck is failing to enforce $SUBJECT.
This is bad, since we aren't really using libpq-fe.h ourselves
in a way that would ensure that c.h symbols don't creep into it.
We can easily do better, as attached, but I wonder which other
headers should get the same treatment.
regards, tom lane
diff --git a/src/tools/pginclude/headerscheck b/src/tools/pginclude/headerscheck
index abbba7aa63..69897092b2 100755
--- a/src/tools/pginclude/headerscheck
+++ b/src/tools/pginclude/headerscheck
@@ -142,7 +142,14 @@ do
# OK, create .c file to include this .h file.
{
- test "$f" != src/include/postgres_fe.h && echo '#include "postgres.h"'
+ # Ideally we'd pre-include only the appropriate one of
+ # postgres.h, postgres_fe.h, or c.h, but we don't have enough
+ # info here to guess what to do in most cases.
+ if test "$f" != src/include/postgres_fe.h -a \
+ "$f" != src/interfaces/libpq/libpq-fe.h
+ then
+ echo '#include "postgres.h"'
+ fi
echo "#include \"$f\""
} >$tmp/test.c