On 11/13/26 12:24 PM, Rainer Jung wrote:
> Hi there,
>
> if a uri string does not contain an explicit port, apr_uri_parse does not add
> the default port from the scheme.
>
> Although uri/apr_uri.c contains the line
>
> 893 uptr->port = apr_uri_port_of_scheme(uptr->scheme);
>
> it doesn't reach it when the port is missing:
>
> 872 if (s == NULL) {
> 873 /* we expect the common case to have no port */
> 874 uptr->hostname = apr_pstrmemdup(p,
> 875 hostinfo + v6_offset1,
> 876 uri - hostinfo - v6_offset2);
> 877 goto deal_with_path;
> 878 }
>
> So it jumps back to the deal_with_path label and the code there always
> returns without getting back to line 893.
>
> Is it desirable for apr_uri_parse to add the default port for a scheme? If
> so, should it be a 1.7.x+ patch only or would that be a
I think this makes kind of sense. If I read the code correctly
http://somewhere:/
would give you port 80
http://somewhere/
would not set the port. This sounds wrong. For reconstructing a URL from an
apr_uri_t apr_uri_unparse does not add a port
explicitly if the port is 0 or if the port is the same as the default port for
this scheme. Hence at least for code
that reconstructs an URL through apr_uri_unparse setting the default port in
apr_uri_parse would not change this.
Other reconstruction code might act differently.
There is another difference between the two case above that might be a bug or
not:
The first case returns an empty string for port_str. In the second case
port_str is NULL.
I am not sure if it makes sense to keep this behavior to indicate both
different cases in the original URL.
Another topic is whether we should set port_str to the default port in case no
port was specified. I think the current behavior of
not doing this is good as it reflects what was found in the URL string.
> valid backport to 1.6.x as well?
I would tink so.
>
> I can create a PR if this seems desirable.
You can, but APR is still on Subversion :-)
Regards
Rüdiger