You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Karl Fogel <kf...@newton.ch.collab.net> on 2003/03/17 22:34:35 UTC
problem interrupting 'svn log' (may be a Neon issue)
kfogel@tigris.org writes:
> Log:
> Now that we handle sigint, 'svn log' needs to explicitly pay attention
> to cancellation. This is an addendum to issue #1004.
There's still a problem here, though, a rather mysterious one.
If you type ^C while running 'svn log' on the whole Subversion
repository, the output will now stop immediately. But your prompt
won't come back immediately -- instead, svn will wait until all the
network data has been received, and *then* give you your prompt back.
(At least, that's what my network activity light seems to be
indicating. :-) )
At first, I thought I knew how to solve this:
"Aha, I just have to go into the RA layers and make sure that when
the `log_receiver' func returns SVN_ERR_CANCELLED, the receiver's
caller stops parsing incoming data and returns the same error
immediately."
But ra_dav should *already* be doing that. Specifically, in
libsvn_ra_dav/log.c:log_end_element(), we store the receiver's error
in a baton and then return NE_XML_INVALID. Since neon/src/ne_xml.h
says this
NE_XML_INVALID ->
No, this is NOT valid XML, and parsing should stop.
...I expect libsvn_ra_dav/util.c:svn_ra_dav__parsed_request() to
return immediately.
Admittedly, the full context of that comment is:
/* Function to check element context...
This callback must return:
NE_XML_VALID ->
Yes, this is valid XML, and I want to handle this element.
NE_XML_INVALID ->
No, this is NOT valid XML, and parsing should stop.
NE_XML_DECLINE ->
I don't know anything about this element, someone else
can handle it.
*/
... and it's not quite clear what function is being referred to. But
anyway, my understanding of NE_XML_INVALID has always been that it
stops the parsing right away.
However, since ra_dav appears to be eating all the rest of the data
from the server before returning to the client library, I guess
parsing isn't stopping?
I can debug this further, but I'm posting early to see if it rings
bells for anyone...
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: problem interrupting 'svn log' (may be a Neon issue)
Posted by Joe Orton <jo...@manyfish.co.uk>.
On Mon, Mar 17, 2003 at 04:34:35PM -0600, Karl Fogel wrote:
> kfogel@tigris.org writes:
> > Log:
> > Now that we handle sigint, 'svn log' needs to explicitly pay attention
> > to cancellation. This is an addendum to issue #1004.
>
> There's still a problem here, though, a rather mysterious one.
>
> If you type ^C while running 'svn log' on the whole Subversion
> repository, the output will now stop immediately. But your prompt
> won't come back immediately -- instead, svn will wait until all the
> network data has been received, and *then* give you your prompt back.
> (At least, that's what my network activity light seems to be
> indicating. :-) )
Yes, neon still eats the whole response regardless - I'm adding an
interface to abort responses from the callbacks for a future version.
(You can do this already if you use the "caller-pulls" interface for
reading response bodies, but not if you use the callback-based
interface).
Regards,
joe
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org