[petsc-dev] testing before merging to master

[Matthew Knepley]

On Sun, Nov 24, 2013 at 6:02 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:

>
> On Nov 24, 2013, at 2:47 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>
> > Barry Smith <bsmith at mcs.anl.gov> writes:
> >>  Jed and I may disagree on this but I believe that if your branch is
> >>
> >> 1) done (that is would be useful for users)
> >>
> >> 2) is completely clean in next
> >>
> >> 3) satisfies PETSc coding standards
> >>
> >> then it should be merged into master and not just “hang around” in
> >> next for days.
> >
> > My reservation is that if you are not confident about the test suite
> > being complete,
>
>     Hanging around in next is not going to make a test suite magically
> become better!!!!  So I would put that in the category of 1) the branch is
> not done because part of its test suite is missing (note that the test
> suite is suppose to be part of a complete branch).
>
> > or if a corresponding change for a dependent package
> > (like petsc4py) is in the works,
>
>    Again 1) the branch is not complete.
>
> >>
> >> then anyone could easily keep track of the two types of merges and
> >> they could be handled properly without requiring someone to remember
> >> something. Is there anyway to do this?
> >
> > How about if the person merging to 'next' pastes the TODO list for the
> > incomplete branch into the commit message.
>
>   Ok, let’s formalize this and you tell us how to automate finding the
> “TODO” list for these branches:
>
>    1) What is the format of the “TODO” list?  For example
> “Todo-before-merge-to-master: ….” ?
>
>    2) What git command do I use to list all branches merged into next (but
> not master) with todo lists and what is in each todo’s list?
>
>       For example, I run   gitmyaliasforfindingincompletebranchesinnext
>  and it produces
>
>            barry/cool-new-newthing : todo - test suite that handles 5
> dimensional case
>            satish/a-feature : todo - provide options database options for
> this feature
>            jed/c-feature : todo - use feature in SLEPc
>
>    Now we have no way of tracking all this stuff and that is bad.


I think we need a better automated way to notify people that checked in
something that a build failed since
there a bunch of builds to check. Can we get the branches merged since the
last build and on any failure/warning
we send the brnach owner an email?

   Matt


>
>   Barry
>

-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20131124/7e744407/attachment.html>