On Mon, 5 Nov 2012, Nick Zitzmann wrote:

>> If nobody objects, I'll merge this into master within shortly.

> I don't know if it re-introduces the original bug that the patch that caused
> the regression fixed or not, but I can verify that, at least when using the
> command line tool with the --anyauth flag, the regression is fixed.

The original bugfix (that introduced the regression) was verified by a set of
new tests provided by Joe Mason. When I reverted the commit, the new tests
started failing and that's why my little patch set also contains an updated
code fix and adjusted test cases.

I believe we're fine now.

> So there are no objections here! The servers our apps deal with generally
> pick a single authentication method for all paths and stick with that
> method.

Right, the problems Joe fixed showed up when you switch auth method on a
re-used connection and clearly that is not a common use case.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html