Re: [Boost-testing] Empty failure log problem [was: Re: Lots of spurious Borland failures?]

From: Vladimir Prus (ghost_at_[hidden])
Date: 2006-11-20 14:12:59

Andreas Huber wrote:

> John Maddock wrote:
>> Guys, we really must get the regressions down to zero: at present
>> there appear to be a lot of spurious failures, for example I'm seeing
>> several regex failures for Borland-5.8.2, but the tests all pass for
>> me locally. What's more the online links all lead to empty failure
>> logs, for example: http://tinyurl.com/u75wu.  Anyone any ideas?
> 
> I'm also suffering from the empty failure log problem. Interestingly,
> all of them are generated by JustSoftwareSolutions-V2. For example, see
> <http://tinyurl.com/y5zedl>...
> 

This suggests that the regression reporting system should also collect full
unprocessed logs from testers, so that such problems can be diagnosed.

- Volodya