Hello,

On Friday 07 November 2008 16:30, Trent W. Buck wrote:
> Eric Kow <[EMAIL PROTECTED]> writes:
> 
> > On Thu, Nov 06, 2008 at 18:43:00 -0800, Trent W. Buck wrote:
> >> Ignore the two-patch bundle I just sent, this folds both (and two
> >> other fixes) into a single patch.
> >> 
> >> Fri Nov  7 13:42:22 EST 2008  Trent W. Buck <[EMAIL PROTECTED]>
> >>   * Remove dangling .lhs references.
> >>   Sorry about this, folks; it seems my conflict merging had a few bugs.
> 
> This kind of cock-up seems to be common (at least for me); often minutes
> after sending a bundle to the list, I'll realize I made a dumb mistake,
> and amend and re-send it.
> 
> To avoid inadvertently applying a cock-up bundle to unstable (which
> would annoy the amend-recording author), how do you (Eric) feel about a
> general rule of not applying (to unstable) any bundle sent in the last
> twenty-four hours?
> 
> That is, after I send a bundle, I have a guaranteed minimum twenty-four
> hour window in which to discover (and amend, resend) cock-ups, and other
> denizens have that same window in which to review the patch and scream
> something like "Noooooo!  Don't apply this; doing so will cause the
> heavens to fall and the ground to be rent asunder."

It seems that this sort of arrangement places a significant, additional, 
burden on the maintainer of the target repository, in this case Eric who 
maintains unstable. The way the GHC people are doing this is described in 
http://hackage.haskell.org/trac/ghc/wiki/TestingPatches. The essence of this 
is that any proposed patch that you send should, at a minimum, have survived 
a test that consists of a build and the successful running of (for GHC: Some 
selection of) test cases.

I must admit that I find this admirably sensible: Falure to comply with this 
that subsequently results in a "validate failure" (validate is the GHC script 
that performs the required build and running of selected test cases) is 
considered a significant event that usually gets you mentioned on the 
appropriate mailing list.

If you need a 24-hour delay before anyone else looks at your patches, It seems 
that you may implement such a delay yourself. If you need someone else to 
look at your patches in a draft state, I guess you could send them as such to 
the mailing list and ask for comments.

> ....

Thanks and best regards
Thorkil
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to