On 11/19/2015 02:44 PM, Martin Sebor wrote:
On 11/18/2015 09:26 PM, Jason Merrill wrote:
The rs6000 target was hitting a bootstrap failure due to
-Werror=type-limits. Since warn_tautological_cmp and other warnings
avoid warning if one of the operands comes from a macro, I thought it
would make sense to do that here as well.
The also disables the warning for functions that are shadowed by
macros such as C atomic_load et al. For example, in the program
below. Is that an acceptable compromise or is there a way to avoid
this?
I think it's an acceptable compromise, but see below.
At the same time, the change doesn't suppress the warning in other
cases where I would have expected it to suppress it based on your
description. For instance here:
unsigned short bar (unsigned short x)
{
#define X x
if (x > 0xffff)
Yes, this is missed because the front end doesn't remember the location
of the use of x; that's one of many location tracking issues. David
Malcolm is working on this stuff.
I noticed there is code elsewhere in c-common.c that avoids
issuing the same warning for system headers (that's the code
that responsible for issuing the warning for the second test
case above).
Hmm, it looks like using expansion_point_if_in_system_header might avoid
the first issue you mention.
There is also code in tree-vrp.c that issues it unconditionally
regardless of macros or system headers.
Good to know.
Jason
