[Pkg-samba-maint] Debian local patches and Samba

[Andrew Bartlett]

On Sun, 2022-05-08 at 23:33 +0300, Michael Tokarev wrote:
> 08.05.2022 23:11, Andrew Bartlett wrote:
> > Michael,
> > 
> > Just a note following up from some comments on Bug#971249.
> > 
> > In the past it has been the strong practice in Samba packaging on
> > Debian not to include local patches, except where those have
> > already
> > been accepted upstream (eg backports from master).
> > 
> > The rationale has been that this avoids divergence, forward-porting 
> > and
> > allows upstream to express and experienced judgement regarding the
> > patches.  It also ensures that, at least in combination with git
> > master, the patch passes CI (avoiding unwanted side-effects).
> 
> Yes, this is a very good practice.  I tend to follow the same rules
> in other packages too, for example qemu.  I *removed* a few patches
> which made debian version to behave differently from upstream in a
> few packages, and already asked about similar patch in debian samba
> too (flipping usershare default).

I think we may be missing some mail from you.  I can't see that mail on
samba-technical.  I think this was quite deliberately changed - the
feature is provided by upstream, if a distribution chooses to enable it
by default to make things 'just work' in Gnome then that's great, we
just can't change that for everyone. 

> What I do there is to move things which by definition do not belong
> to where they're put, and which does not affect anything else.
> Like the socket directory - it must not be in /var, there's /run for
> it.

I have seen this mail.  I don't have any strong views on that, it looks
like an accident of history, but the next step is a GitLab MR so it can
be checked via CI.  

> Another case is the spelling fixes I posted initially. 

I can't see that patch on samba-technical, nor have I seen patches from
you on GitLab.  

It seems we are at an impasse per
https://bugzilla.samba.org/show_bug.cgi?id=15039#c6 

We specifically request that patches not be submitted to Bugzilla
initially. 

> I do not want
> to remove this patch from debian patches, even if upstream does not
> want to accept it (tho I should clean it up to remove places where
> it touches the published docs).  I can also remove the place where
> it renames a local variable, - this is something which touches the
> code. Hopefully there's no tests needs to be written for the spelling
> fixes like this.
> 
> The change in testparm (to stop it from erroring out if /run/samba
> does not exist) is another example of "obvious-wrong", and it lets
> the packaging to have much less stupid workarounds. If that's not
> accepted upstream we can revert it, - it wont affect anything but
> our own startup scripts and may introduce the old bugs again.

Again, if that is inn Bugzilla then we probably have the answer.

> > I think this has served us well, and I would continue to encourage
> > the
> > submission of patches upstream via GitLab per
> > https://wiki.samba.org/index.php/Contribute
> > 
> > If you have any difficulties with this process please discuss on
> > samba-
> > technical, we remain glad to help ensure Samba is as good as
> > possible
> > for Debian.
> 
> Well, I posted several questions and proposals to samba-techincal@,
> none of which were answered. Including all recent changes.

I've been on leave myself, but I've caught up on some of your recent
questions. 

> I don't plan to add more stuff.  Actually I didn't plan to add any
> stuff at all, I just wanted to fix a bug :)

Thanks,

Andrew Bartlett

-- 
Andrew Bartlett (he/him)       https://samba.org/~abartlet/
Samba Team Member (since 2001) https://samba.org
Samba Team Lead, Catalyst IT   https://catalyst.net.nz/services/samba

Samba Development and Support, Catalyst IT - Expert Open Source
Solutions