Heimdal Patches Re: [PATCH] Some coverity fixes

[Andrew Bartlett]

On Mon, 2015-05-11 at 08:59 +0200, Volker Lendecke wrote:
> On Sat, May 09, 2015 at 08:07:38AM +0100, Andrew Bartlett wrote:
> > Actually, they really do hurt.  I am working on updating Heimdal (it is
> > a long, thankless task, but still clearly required), I've had to drop
> > almost all of the 'fix the build' patches so far, because they simply
> > don't apply.  
> 
> Does this mean we will have to go through a load of build failures
> again? 

Yes, I expect that will be required.  That is why it would be really
helpful if you could do that work upstream, where they can be included
in the next release. 

I do value your passion for improving Samba, and our build, I just need
you to direct your energies in a particular path, so we all get a great
outcome here.

> We seem to have much stricter warning levels than heimdal has,
> and I do see many failures over and over again with the O3 developer
> build. This also is painful.
> 
> I request to instrument waf such that in the heimdal tree we allow every
> warning under the sun and not fail at all. Please see a new heimdal
> import NACKed by me already if this is not met.

Waf is certainly flexible enough to turn off the stricter build settings
per-module, if we need to.  Even better however would be to just work
with the upstream maintainers of this third party code to improve it for
all users.

> Also, I will modify our Coverity build such that we completely ignore
> errors there.

Heimdal is using Coverity upstream, but I don't see why you would need
to ignore the errors.  Does the coverity report have a limitation where
this stops you doing other work?

> > To ensure we don't just loose them when we rebase (and the rebase is
> > going to be so large you would never notice), I really do need patches
> > to Heimdal to be first submitted as github pull requests into Heimdal
> > master.  We have tools to try and pick up the local patches, but they
> > are no longer very helpful because we have a mix of patches that do
> > apply, don't apply upstream divergence or are already upstream.
> 
> Okay, will heimdal maintain a release stream based on our version for us,
> with a latency measured in months not years? Much like Michael maintains
> ctdb-1.0.114 forever?

I'm not sure of the details of ctdb-1.0.114, but I don't, based on the
small number of patches outstanding, see a need for a Samba specific
Heimdal release stream.  The number of outstanding patches is quite
small, and if we keep up to date regularly, and if you do the work to
have Heimdal build with your preferred combinations of -O3 and -Werror
upstream, then the imports should be much less trouble.

Love is keen to have us use a released upstream version, and if
supported and encouraged by the team, I'm happy to continue working
towards that goal. 

Thanks!

Andrew Bartlett

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba