[darcs-users] Re: first impressions on a 'darcs send' to a special email address workflow

[Mark Stosberg]

On 2005-03-03, Thomas Zander <zander at kde.org> wrote:
>> >
>> > Note that you can run the tests when doing darcs push
>>
>> Can you have the best both worlds: Push the patches, but get an e-mail
>> with a result after the tests have run?
>
> That hardly ever makes sense; if you want to do post-commit testing (i.e.=20
> don't revert the changes if the tests fail) simply setup your server to do=
>=20
> nightly test runs.
> With your suggestion you have to assume that the person who pushed the patc=
> h=20
> will read the email and unrecord/amend the local patch and not just push it=
>=20
> again.  That just does not give you a very good workflow IMO.

That's an interesting way to look at. I was thinking: 

"It would be like using darcs send, without the overhead of setting up the email bits."

The usage model that we've settled into here is "none of the above". 

Our code comprises a mod_perl enabled website.

We have a 'smokebot' that runs twice a day on our 'alpha' central repo,
and e-mails the developers if one of our 4,000 tests fails.

This means we only find out about some failures after the fact, but in
our work flow, that tends to be good enough. 

The standard darcs testing infrastructure would be a headache, because
it means configuring a mod_perl server on the fly to work with a
temporary directory that we don't know the name of. 

It's also nice to have tests run asynchronously so we don't have to wait
for them. 

If it wasn't the for the smokebot, the whole test suite would get run
much less frequently. 

    Mark