tdf#96482 (5.0.4 regression introduced in fix of tdf#94069) [was: minutes of ESC call ...]

[Lionel Elie Mamane]

On Mon, Jan 25, 2016 at 05:28:53PM +0000, Michael Meeks wrote:
1;2802;0c> 
> On Mon, 2016-01-25 at 16:49 +0000, Caolán McNamara wrote:
> > So, some bisecting shows that its 
> 
> 	Wow =)

Thanks a lot Caolán for the bisect.

> > commit 87199d3829257420429057336283c55be6ae7481
> > Author: Michael Meeks <michael.meeks at collabora.com>
> > Date:   Tue Nov 24 16:59:29 2015 +0000
> > 
> >     vcl: re-introduce idle handling.
> > 
> > which fixed the problem on master/5-1. Its non-trivial to backport that
> > though to 5-0. Not sure how that should look if applied to 5-0
> 
> 	Right; it's ~impossible to back-port without some seriously
> heavy-lifting / re-qualification process - I'm amused/interested to hear
> that it fixed something ;-) I'd not support it going into -5-0 really.

> 	I'd suggest that it'd be better to work out what two (or more) events
> are getting mis-ordered to cause the issue in the bug - and get stack
> traces from those code-points in both 5.0 and 5.1 - and I'd be happy to
> take a look; perhaps in-person at FOSDEM.

You are saying that this commit is not in 5-1? That's interesting
because the bug is not present in 5-1. So that means that yet another
commit "corrected" the bug in 5-1,

I'd be happy to do the stack trace stuff with you during the
Hackfest-before-FOSDEM.

-- 
Lionel