Re: [PATCH] avoid folding snprintf calls with bounds > INT_MAX (PR 87096)

[Jeff Law <law@redhat.com>]

On 12/12/18 4:18 PM, Martin Sebor wrote:
> POSIX requires snprintf to fail with EOVERFLOW when the specified
> bound exceeds INT_MAX.  This requirement conflicts with the C
> requirements on valid calls to the function and isn't universally
> implemented (e.g., Glibc doesn't seem to follow it, and GCC has
> historically not paid heed to it either).  Nevertheless, there
> are implementations that do respect it (Solaris being one of
> them), and it seems that GCC should make a tricky situation
> even more treacherous for programmers by returning different
> values from some calls to the function than the library would.
> This is also what bug 87096 requests.  The patch also adds
> a warning that was missing from a subset of these troublesome
> calls.
> 
> The attached patch disables the snprintf constant folding and
> range optimization for calls to it and other related bounded
> functions when the bound is not known not to exceed INT_MAX.
> 
> Tested on x86_64-linux.
> 
> Martin
> 
> gcc-87096.diff
> 
> PR tree-optimization/87096 - Optimised snprintf is not POSIX conformant
> 
> gcc/ChangeLog:
> 
> 	PR rtl-optimization/87096
> 	* gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid
> 	folding calls whose bound may exceed INT_MAX.  Diagnose bound ranges
> 	that exceed the limit.
> 
> gcc/testsuite/ChangeLog:
> 
> 	PR tree-optimization/87096
> 	* gcc.dg/tree-ssa/builtin-snprintf-4.c: New test.
OK
jeff