Auckland Testing Summary 15 November 2012

[Tom Parker]

On 16/12/12 17:20, Mikus Grinbergs wrote:

> I believe the symptom has to do with the presence of multiple windows by
> an Activity.  Sugar starts the initial window - if in time certain
> things do not happen in that window, the "failed to start" message
> eventually gets displayed.  But if in the meantime the Activity is
> proceeding in ANOTHER window (e.g., screen) , the eventual message in
> the initial window is misleading.

Then why do we see this error seemingly randomly?  If the activity isn't 
doing things on it's initial window, then wouldn't it always trigger 
this bug?

Does the shell keep track of the launched process? There are ways for an 
activity to die immediately on startup but leave the starting screen 
pulsing for ages followed by a failed to start screen. This would be 
improved if the shell noticed the launched process was died immediately.

Would it help with debugging if it dropped information about what is 
going on when the failed to start warning is triggered? Something like 
"launched ... got a & b, but never saw c after n seconds, child process 
... still running, backtrace of that process follows".

Currently all we can tell you is "... triggered the failed to start but 
otherwise worked normally" which is apparently insufficient to chase all 
the bugs out of the system.