> On Apr 29, 2015, at 9:43 PM, Dmitry Karpeyev <[email protected]> wrote: > > MatConvert_Nest_AIJ() hasn't really been tested enough and is (evidently) > buggy. > This particular use has been commented out in master,
Commented out? Is that how we manage code? > but is turned on in next. > We need more elementary tests that would help debug this routine. > I'll try to get to it in the next week or so, but maybe comment this back out > in the meantime? I think one should take whatever branch that is badly broken and produces these errors out of next (since that branch is clearly not good), sadly I don't know how to do that with git and Jed could spend an hour explaining rationally why it should be impossible in git to remove that branch from next. So my response is to stubbornly say "just remove that bad branch from next, and if you can't do that with git branches than git branches need work". Note that it is not good to have things permanently broken in next because that makes it difficult to promote any other branch from next to master since it is difficult to know if that other branch would pass all the tests if the broken branch was not in next. Barry > > Dmitry. > > On Wed, Apr 29, 2015 at 3:44 PM Barry Smith <[email protected]> wrote: > > Whoever put the MatConvert_Nest_AIJ() code into next please fix in your > branch and update next > > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2015/04/29/examples_next_arch-freebsd-cxx-pkgs-opt_wii.log > > Thanks > > >
