bug#72652: 31.0.50; url-retrieve on non-existent domain gives no indication of error

[Eli Zaretskii <eliz@gnu.org>]

> cc: 72652@debbugs.gnu.org
> From: Greg Minshall <minshall@umich.edu>
> Comments: In-reply-to Eli Zaretskii <eliz@gnu.org>
>    message dated "Fri, 16 Aug 2024 10:43:30 +0300."
> Date: Fri, 16 Aug 2024 22:39:11 +0300
> 
> > On MS-Windows (mytry 2) signals an error:
> 
> i see.  possibly the Emacs on windows is using a synchronous DNS
> resolver (possibly not).

Yes, I think so.

> the problem is that there's no indication from url-retrieve that the
> requested transfer has failed (or has in any way finished).  the only
> reason the message about the buffer was printed out was because the
> mytry script times out after 5 seconds, prints out whatever might be in
> the buffer.
> 
> so, a program that calls url-retrieve for an unknown host is never
> notified that anything happened.
> 
> 
> if some sort of error is detected synchronous with the call to
> url-retrieve (i.e., before url-retrieve returns to its caller),
> signaling an error seems fine.
> 
> in the case where the error is detected later, then i think (if
> possible) url-retrieve should cause a callback to happen, with the
> `:error` data in the status parameter specifying something like "unknown
> host".
> 
> in the case of (mytry 1), there is a callback (to the callback routine
> specified in the invocation of url-retrieve) which passes an error
> indication:
> ----
> called back: ((:error (error http 500)
> ----
> (one can modify mytry to print out the status given to the
> callback routine -- see below.)

Patches to report an error in the (mytry 2) case are welcome.