[cfe-commits] r135075 - in /cfe/trunk: lib/Analysis/FormatString.cpp test/Sema/format-strings-i386.c

[Joerg Sonnenberger]

On Wed, Jul 13, 2011 at 08:20:58PM -0000, Ted Kremenek wrote:
> Author: kremenek
> Date: Wed Jul 13 15:20:58 2011
> New Revision: 135075
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=135075&view=rev
> Log:
> format string checking: long and int have the same widths on 32-bit, so we shouldn't warn about using
> an "int" format specifier with a "long" type in 32-bit.

I don't like this change. I can think of the following cases for why
this warning is triggered and useful:

(1) Mismatch between PRI* and (s)size_t / ptrdiff_t definitions. This
was a real issue on PPC for FreeBSD and NetBSD.

(2) Size assumptions about (u)int32_t and it getting defined as long.

(3) Code that breaks on 64bit builds or uses different #if branches in
that case. This still sounds like a valid bug.

So I thing it should be reverted.

Joerg