There are quite a few places in the pjsip code where it is assumed that an
snprintf of "%.*s" will nicely print nothing if the supplied arguments are 0
and NULL.
For example, in pjsua_acc.c there is:
len = pj_ansi_snprintf(tmp, PJSIP_MAX_URL_SIZE,
"<sip:%.*s%s%s%.*s%s:%d;transport=%s%.*s>%.*s",
(int)acc->user_part.slen,
acc->user_part.ptr,
(acc->user_part.slen? "@" : ""),
beginquote,
(int)via_addr->slen,
via_addr->ptr,
endquote,
rport,
tp->type_name,
(int)acc->cfg.contact_uri_params.slen,
acc->cfg.contact_uri_params.ptr,
(int)acc->cfg.contact_params.slen,
acc->cfg.contact_params.ptr);
... and by default acc->cfg.contact_uri_params and acc->cfg.contact_params
both have 0 lengths and NULL pointers.
This is very convenient when it works but sadly is not a feature of C
libraries that one can count on, which makes the code less than portable.
In my case, i'm using a version of uclibc that prints "(null)" whenever the
pointer is NULL, regardless of the "precision" argument.
Modifying uclibc is a possible work around for me, although it means i have to
fork a version of uclibc which will make it hard to track their latest
version... which won't make people i work with happy.
I suggest that all such cases in the pj code be slightly rewritten so that ""
is substituted where "ptr" is NULL.
This can best be done with a wrapper inline function which we already have in
pj/string.h !!! :
PJ_INLINE(const char*) pj_strbuf( const pj_str_t *str )
{
return str->ptr;
}
used as:
pj_strbuf(&acc->cfg.contact_params)
To generalize, it would be better if the fields of pj_str_t were accessed only
by standard access functions pj_strbuf() and pj_strlen() already provided in
pj/string.h , and never directly as is the case in so much of the pj code...
however, fixing the snprintfs would be a good start.
Thanks a lot for pjsip,
-Ted Merrill
