Il 02/10/2012 13:06, Juan Quintela ha scritto:
>> Perhaps the other way round:
>> >
>> > if (ret < 0) {
>> > ret2 = -1;
>> > }
>> > ...
>> > return ret2;
> This lost the 1st errno value.
Right, I meant ret2 = ret;
> The other way around we preserve it.
I think with "ret2 = ret;" it is the same.
The way you have it in your patch, you could truncate a positive ssize_t
return value from buffered_flush that does not fit in an int.
Theoretical, I know, but I can see Coverity spotting it from a mile...
Paolo
