dulwich-users team mailing list archive

[Jelmer Vernooij <jelmer@xxxxxxxxx>]

On Sun, 2010-12-19 at 13:57 -0800, Dave Borowitz wrote:
> AFAIK nose is still canonical, since it's what's in the Makefile. I'm
> fine with this provided:
> -if no git is installed, tests are still skipped rather than failed
> -you rewrite the 'check' make target such that it doesn't run the
> non-compat tests twice.
FWIW this is how trunk behaves at the moment. 

> A nice-to-have would be a 'check-nocompat' or similar make target. I
> for one do things like run the tests on a bunch of sequential patches,
> and it's nice when those take 1s each rather than 10s.
Adding a check-nocompat target seems reasonable to me. 

I use testr for most of my projects, it can e.g. parallelize test runs
and re-run only the failing tests from the previous run. 

% time testr run --parallel
...
id=59, tests=423, skips=6
testr run --parallel  0.64s user 0.22s system 18% cpu 4.776 total

Cheers,

Jelmer

> On Sun, Dec 19, 2010 at 13:36, Augie Fackler <durin42@xxxxxxxxx>
> wrote:
>         https://github.com/jelmer/dulwich/commit/b5490da68052e33b904e32c04f2aef140c8bcb45
>         
>         This means that nose will no longer skip these tests by
>         default, which we had historically (as I understood things,
>         anyway) wanted. Is nose no longer the canonical way of running
>         tests? I thought we didn't want to support test runners
>         without discovery support? (I thought unittest2 supported
>         discovery...)
>         
>         Thanks,
>         Augie
>         
>         
>         _______________________________________________
>         Mailing list: https://launchpad.net/~dulwich-users
>         Post to     : dulwich-users@xxxxxxxxxxxxxxxxxxx
>         Unsubscribe : https://launchpad.net/~dulwich-users
>         More help   : https://help.launchpad.net/ListHelp
> 
>

Attachment: signature.asc
Description: This is a digitally signed message part