On Fri, Jul 23, 2010 at 11:51:55AM -0700, [email protected] wrote:
> On Thu, Jul 22, 2010 at 05:28:43PM -0700, Shawn Walker wrote:
> > On 07/22/10 03:30 PM, [email protected] wrote:
> > >As part of my putback for updating the publication tools, I removed the
> > >ability for publication clients to read HTTP response bodies when they
> > >recieve an error.  Unfortunately, this is the only way for the depot to
> > >provide more details about a publication failure.  I had hoped the
> > >removal wouldn't be noticed, but it was, and the functionality is still
> > >needed.
> > >
> > >This webrev now retains the response body for publication operations
> > >that fail with HTTP 400.  The additional information is printed when the
> > >exception is turned into a string.
> > >
> > >   http://cr.opensolaris.org/~johansen/webrev-16607/
> > 
> > I tested this locally against gate tip and it worked as expected.
> > This is also a lot nicer than it used to be.
> 
> Thanks for looking at this.  It took more testing than I expected.
> Aside from the changes to the fileobject to allow the caller to access
> the response body in an error, I also had to change the engine to
> continue in some error cases.  Previously, if a caller specified a
> write-callback, the engine instructed CURL to stop immediately if a HTTP
> error >= 400 was seen.  I had to switch this off to get the response
> body for the fileobj.
> 
> After thinking about this some more, I'm wondering if it would actually
> be better to have all transactions as FAILONERROR by default, since that
> would prevent other errors from downloading response bodies too.  Agree?

I went ahead and implemented the change to use FAILONERROR more
generally.  This webrev is the changes between the previous webrev and
the current workspace.

http://cr.opensolaris.org/~johansen/webrev-16607-2-1/

-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to