Christian Grothoff píše v Čt 10. 01. 2013 v 10:57 +0100: > On 01/10/2013 10:40 AM, Václav Pavlín wrote: > There is also again the question of how to possibly handle this > condition (other than by logging). As it is harmless, a log message may > confuse more than help -- after all, things are working. So I actually > consciously decided against doing anything in this particular case. We > have a similar 'setsockopt' case for IPV6_V6ONLY in daemon.c; again, the > failure is "mostly harmless" and actual handling of errors is likely to > do more harm (as in, break stuff) than ignoring the return value. >
Ok, I've found some article about TCP_CORK where was stated that if there is some incomplete packet and you don't remove the cork they won't be sent. But I got your point. > Coverity is making invalid assumptions. "left" is of type "off_t", > which can be 32 bit or 64 bit, > signed or unsigned. SSIZE_MAX is the maximum value of 'ssize_t', which > can be 32-bit or 64 bit signed; and these are actually only the > constraints on _common_ platforms; exotic systems might be even more crazy. > > So on some platforms this check may be unnecessary, but not on all. I tried to go through code to check the types but I probably didn't cover all the branches of IFDEFs. Thanks for explanation. > > Happy hacking! > > Christian > Regards Vaclav Pavlin
