Re: criteria.html open issues

[Neil Booth <neil at daikokuya dot demon dot co dot uk>]

Joseph S. Myers wrote:-

> So, your design will now ensure that there is no need to define
> __LONG_MAX__ and similar macros manually in order to get <limits.h> right,
> whatever the target and command line options?  I rather suspect that every
> macro defined in <limits.h> is wrong for at least one combination at
> present.

No, I don't think so (at least, I'm not sure what getting that right
requires).  It might be possible, though, with sufficient target
information (via macros, or whatever).

I was referring more to stuff like -fhosted and -ffreestanding
conflicting on the command line, or -fchar-signed and -fchar-unsigned.
IIRC in a post a few weeks back you were suggesting we don't define
macros for such things in the driver (since we get it wrong at present
if both appear).  I think we can handle such things in the driver
without problems with a non-SPEC implementation (and maybe change the
front ends to only accept one of the variant forms of -fhosted /
-fno-freestanding, doing a translation in the driver).

Another example is the endless aliases for -std=XXX that CPP accepts.
The driver could alias those down to a minimal subset, reducing the
burden on cpplib for what it needs to accept.

Neil.