> Hi, > > valgrind reports that the socket name is read more than its size, it is > right since it is 64 bytes but 102 are read. > > It looks to me that the rest of the code assumes that 102 is the right > size so use it in malloc() too. > > The attached patch suffice to shut up valgrind. > > Please review and give it some testing. > > How to reproduce:
Those 64 bytes are used by the HTTP embedded server to generate a pseudo-random name for a UNIX socket (if the user does not specify it). 102 instead is a compromise on the various UNIX socket implementations that do not adere to a common value, and is used when the user specify a socket name on the command line. So the check about "102" is done only for the user-supplied sockets name, the http server alloc the minimum required ammount of memory to contains the random socket name + the '0' final byte. -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
