bug#8435: misuse of error ("...%d...", ...) on 64-bit hosts

[Paul Eggert <eggert@cs.ucla.edu>]

On 04/09/2011 01:21 PM, Eli Zaretskii wrote:
>> Date: Sat, 09 Apr 2011 12:39:59 -0700
>> From: Paul Eggert <eggert@cs.ucla.edu>
>>
>> As Emacs evolves, if we find that more lines of code are affected,
>> then the tradeoffs will change.  If that happens, it shouldn't
>> be hard to come up with a doprnt replacement that uses
>> vsnprintf internally and that also handles multibyte character
>> truncation and non-8-bit codepoints.
> 
> I'd prefer that we do this now.

OK, please feel free to do that.  To help move this along, I
resurrected src/doprnt.c in my patch, and merged it into
the trunk, along with all the other patches I've been testing
that have to do with GCC 4.6.0's static checks.  Currently
Emacs is not using src/doprnt.c but it shouldn't be hard to
refactor the code to use doprnt again if that's what you prefer.
This should result in some simplification of vsnprintf's two
callers verror and vmessage.  I still don't think it's worth
the hassle, given Emacs's current usage (but I guess you've
been warned :-).

If you take this project on, you need to fix the 64-bit related
problems in doprnt.  For example, it's not safe to copy
a string length into an 'int'.  I've mentioned other bugs
in this area, and I'm sure there are others that I haven't
mentioned (I gave up on doprnt before fully analyzing it).

I'm going to mark this bug as "done", since the bug itself
is fixed now, and we're now talking about refactoring the fix.