Mozilla/5.0 (X11; Linux x86_64; rv:10.0.1) Gecko/20120209 Thunderbird/10.0.1
On 02/28/2012 03:13 PM, Jim Meyering wrote:
>> -#define OUTPUT_BLOCK_SLOP (page_size - 1)
>> > +#define OUTPUT_BLOCK_SLOP MAX (sizeof (uintptr_t), page_size - 1)
> I haven't seen justification for why you're making the above change.
> Can sizeof uintptr_t really be larger than page_size-1 (getpagesize()-1)?
> This seems so unlikely that it'd deserve an assertion in main where
> page_size is set, even though there are only two uses of OUTPUT_BLOCK_SLOP.
Theoretically getpagesize could return 1, no?
Admittedly it's unlikely. I guess it'd be OK
to omit this change, and to just put in a comment
saying something like "We use the output slop
to store a uintptr_t value, so we assume that
getpagesize returns a value greater than sizeof (uintptr_t)."
Perhaps the comment should go right next to the place where
the uintptr_t sentinel is inserted.
> If you leave the sizeof, please omit the parentheses.
But "sizeof uintptr_t" wouldn't work, as uintptr_t is a type.