Re: gcc unsigned vs signed compare

[Matt Thomas <matt%3am-software.com@localhost>]

On Aug 31, 2011, at 1:56 PM, Christos Zoulas wrote:

> In article <20110831202534.GA7800%antioche.eu.org@localhost>,
> Manuel Bouyer  <bouyer%antioche.eu.org@localhost> wrote:
>> Hello,
>> I tracked down a bug on a mips64el platform (see PR 45310), which turns out
>> to be a signed vs unsigned compare issue.
>> 
>> In lib/libc/gen/fixunsgen_ieee754.c there is a compare:
>>      (exp - mant_dig > sizeof(UINTXX_T)*8-1)
>> exp is int, mant_dig is size_t (unsigned long) and sizeof returns size_t.
>> gcc generates an unsigned compare for this, which doesn't gives the expected
>> result when exp < mant_dig. Is gcc right when generating an unsigned
>> compare here ?
> 
> Yes, the promotion rule is signed -> unsigned.

I fixed this by changing it to

        ((size_t)exp >= mant_dig + sizeof(UINTXX_T)*8)

which also handles the the (exp < 0 || ...) before that.