<danw> it's not clear why they think they need to call that function
<walters> no pwithnall here at the moment
<danw> you could try replacing it with soup_session_cancel_message (priv->session, priv->message, SOUP_STATUS_CANCELLED); and see if "make check" still passes...
<walters> looks like there are pre-existing test failures here
* walters pushes to keep things building and shipping
<danw> walters: i have no idea if that patch will actually work

(In reply to comment #1)
> <danw> it's not clear why they think they need to call that function

That code was originally copied from SoupInputStream, which used soup_message_io_cleanup(). I can’t remember why I didn’t remove the call.

(See: http://git.gnome.org/browse/libgdata/commit/?id=d35e5400)

> <danw> you could try replacing it with soup_session_cancel_message
> (priv->session, priv->message, SOUP_STATUS_CANCELLED); and see if "make check"
> still passes...

The relevant tests are:
 • streams
 • picasaweb: /picasaweb/download/photo, /picasaweb/download/thumbnails
 • documents: /documents/download/document, /documents/download/thumbnail

These all still pass except the first documents one, but that’s a known (pre-existing) failure.

> <walters> looks like there are pre-existing test failures here

Unfortunately there will probably always be test failures in libgdata, unless I magic up the time to rewrite the entire test suite to work against a mock Google server, rather than the real thing. Although, arguably, that would render the tests meaningless whenever the real Google servers’ behaviour diverged from the mock server.

> <danw> walters: i have no idea if that patch will actually work

Looks OK to me, for what my opinion’s worth. Thanks for fixing this up.