https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128429
Alexandre Oliva <aoliva at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |INVALID
--- Comment #4 from Alexandre Oliva <aoliva at gcc dot gnu.org> ---
Ugh. I see, so there are goals at odds with each other.
On the one hand, we want to get warnings during testing that we wouldn't get
from system headers, and we want those warnings not only from libstdc++
developers, but from all testers.
On the other hand, we have error messages that are tested for that currently
depend on the non-system header status of libstdc++ headers, and that produce
lower-qualify (and different from expected) error messages when libstdc++
headers are considered system headers.
I guess, if the compiler will pay attention to the system header status of
intermediate types when simplifying type names for error printing, we could
adjust the tests and get used to finding out, after the fact, that users don't
get the better error messages we get during development testing. That's a risk
that generally comes with testing something different from what users will run,
such as considering libstdc++ headers non-system ones.
Oh well... I don't like it, but if we're doing that deliberately, we'll have
to live with those consequences. Thanks!
