Re: GNU M4 1.4.8b released (beta release)

[Matthew Woehlke]

Paul Eggert wrote:
> Eric Blake writes:
> > I tried finding something in POSIX or C99 that said that an implementation
> > could provide long long without unsigned long long, but did not see
> > anything obvious
>
> C99 requires both types.  C89 requires neither, and Tandem provides just
> one of them as an extension to C89.  That's allowed.
>
> > Therefore, we can argue that Tandem is buggy
>
> No, we could argue that only if we assumed C99, which we don't.
> Maybe in 10 or 15 years we can assume C99, but by then the issue
> will be moot, as newer Tandem releases support both.

Actually, I'm inclined to agree with Eric's reading of POSIX: "If an 
implementation provides integer types with width 64 that meet these 
requirements, then the following types are required: int64_t uint64_t".

In m4's case I'm inclined to agree that not supporting 'long long' at 
all seems to be the way to go. In the long run, I'm still not convinced 
this won't come back to bite something that relies more heavily on 
file-system calls (since some system calls still return 64-bit data types).

I could always try to resurrect the gnulib printf patch? (But I probably 
wouldn't vote for trying to shoehorn it into m4-1.4.9.)

--
Matthew
"Have you tried that new mixed drink, 'GDR'"?
"What is it?"
"Gin, Duck and Rum. It tastes fowl."