[petsc-dev] MatMatMult gives different results

[Satish Balay]

On Thu, 9 Feb 2012, Jed Brown wrote:

> On Thu, Feb 9, 2012 at 20:29, Satish Balay <balay at mcs.anl.gov> wrote:
> 
> > Is it possible to add such a trigger to bitbucket?
> >
> 
> Yes
> 
> http://confluence.atlassian.com/display/BITBUCKET/Setting+Up+the+Bitbucket+POST+Service

unless I'm mistaken - it looks like https - not ssh.

> > > The simple convenience of having https:// and
> > > ssh:// trivially push/pull the same repository is enough to make me
> > prefer
> > > interacting with bitbucket (screwy networks filter ssh).
> >
> > Not sure what problems you had. The only known issue with ssh is - we
> > have fail2ban running on petsc.cs.iit - and when there are ssh
> > connections from misconfigured ip adresses [where dns lookup does not
> > work - it thinks its a breakin attempt - and tries to ban that
> > access].
> >
> 
> No, the problem is goofy wireless networks. This bit Matt when he was at
> Brown a few weeks ago

one of matts issues was what I described [don't know if it was from
brown or somewhere else]

> and me in Istanbul last week. AFAIK, we have had to
> decide for each repo whether to use ssh or https for pushing, so I can't hg
> push https://petsc.cs.iit.edu/petsc/petsc-dev. Bitbucket supports both
> simultaneously, and the user manages their credentials so you don't have to.

yes - I'd rather not do https on petsc.cs.iit.edu [I could - but as
you say - its combersome compared to bitbucket].

I guess the alternative - is to push to your personal bitbucket clone
- and ask someone else to pull/sync to the master. [Isn't this one of
the things you and Sean were advocating as advantages of bitbucket? If
so - I think it should be equally useful for syncing with petsc-dev -
when hosted at petsc.cs.iit.edu]

Satish