|
|
Boost Testing : |
From: Vladimir Prus (ghost_at_[hidden])
Date: 2007-01-23 02:25:37
Rene Rivera wrote:
> Vladimir Prus wrote:
>> Roland Schwarz wrote:
>>
>>>> I feel that if a library author is willing to devote effort to backward
>>>> compatibility he should rather be encouraged than hampered.
>>> Humm, yes. But why put the load on the regression runners then?
>>> Regression runners mainly contribute CPU cycles, not development. Isn't
>>> this the duty of the library author to make everything as smooth as
>>> possible? Preparing a suitable run of regressions is part of these
>>> duties. Not?
>>
>> I think it would be good to have a scheme where all possible setup
>> for regression runners, except for the set of installed compilers,
>> is centrally managed, so that regression runners won't ever
>> have to tweak anything at all. Or better yet -- won't be able
>> to change anything.
>>
>> Such scheme does not exist at the moment, unfortunately.
>
> That's how BuildBot works.
I'm afraid that's irrelevant -- we don't use BuildBot.
> The master machine has the configuration for
> all the bots which it transmits to each to control what and how they run
> actions.
And I did not realize BuildBot has this nice property until
Stefan Seefeld told me that a few days ago ;-)
> PS. To answer a much earlier question about whatever happened to using
> buildbot for Boost. I've been swamped since last Summer, and the machine
> that was my Linux server only came back online recently. So I haven't
> the free time to work on it.
Understood. I don't know how to have it done, but we need a scheme were
it's extra hard for an individual regression runner to make mistake
that will result in 1000 failures being reported. Post 1.34, sigh.
- Volodya