On Tue, 23 Dec 2008 15:43:46 -0800
Roland Dreier <rdre...@cisco.com> wrote:
> > Also that is in the test code (libibnetdisc/test/iblinkinfotest.c). The
> real
> > iblinkinfo has them all declared static.
> >
> > void
> > print_port(ibnd_node_t *node, ibnd_port_t *port)
> > {
> > static char remote_guid_str[256];
> > static char remote_str[256];
>
> so this function is not thread-safe.... This seems to be setting an
> unnecessary booby-trap; is there a reason not to put these variables on
> the stack?
Just what Sean mentioned.
There are actually 2 print_port functions in the 3 patches which were sent.
One is in a prototype tool called iblinkinfotest which was left in the library
directory and optionally built for testing. The other is in a single threaded
tool which uses the library maintained in the official src dir of
infiniband_diags. _Neither_ function _is_ part of the library and both are in
single threaded tools.
Ira
_______________________________________________
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
