Spec files maintained in external source control: *please* mention this in the spec file

[Adam Williamson]

Hi folks! So I got bitten again today by the situation where the
primary contact for a given package considers the 'canonical' source
for the spec file to be some external SCM, and finds it a problem when
someone (e.g. a provenpackager like me...) changes the package directly
in dist-git.

This is at least 50% my fault for not trying to check in before
changing the package, but at the same time, it'd really be helpful if
packagers who do this could note it clearly and prominently in the spec
file, with a direct link to the external repo in question. Fedora dist-
git usually is supposed to be the canonical location for Fedora package
files, but if you're following a different workflow, it'd really help
avoid problems if you at least explain this prominently in such a way
that someone who just goes directly to dist-git will be aware of it.
It's been the case for a while now that Fedora is supposed to be
broadly collaborative and no-one has exclusive 'ownership' of a package
any more (hence the name change to 'primary contact' from 'maintainer'
or 'package owner'), so I think it's reasonable to expect this.

Thanks!
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net

[Zbigniew Jędrzejewski-Szmek]

On Mon, Feb 13, 2017 at 09:44:50AM -0800, Adam Williamson wrote:
>  Hi folks! So I got bitten again today by the situation where the
>  primary contact for a given package considers the 'canonical' source
>  for the spec file to be some external SCM, and finds it a problem when
>  someone (e.g. a provenpackager like me...) changes the package directly
>  in dist-git. 
https://fedorahosted.org/fpc/ticket/613
It's seems to be going nowhere.

>  This is at least 50% my fault for not trying to check in before
>  changing the package 
It'd help if there was a standard way to mark such
things.

Zbyszek

[Mat Booth]

On 14 February 2017 at 02:58, Zbigniew Jędrzejewski-Szmek <zbyszek(a)in.waw.pl
>  wrote: 
>  On Mon, Feb 13, 2017 at 09:44:50AM -0800, Adam Williamson wrote:
>  > Hi folks! So I got bitten again today by the situation where the
>  > primary contact for a given package considers the 'canonical' source
>  > for the spec file to be some external SCM, and finds it a problem when
>  > someone (e.g. a provenpackager like me...) changes the package directly
>  > in dist-git.
>
>  https://fedorahosted.org/fpc/ticket/613
>  It's seems to be going nowhere.
>
>  > This is at least 50% my fault for not trying to check in before
>  > changing the package
>  It'd help if there was a standard way to mark such things.
>
>  Zbyszek
>  _______________________________________________
>  devel mailing list -- devel(a)lists.fedoraproject.org
>  To unsubscribe send an email to devel-leave(a)lists.fedoraproject.org
>  

Ticket is still has "need info" status  -- if the info is provided, the
status should be set to "discuss at next meeting"

-- 
Mat Booth
http://fedoraproject.org/get-fedora

[Zbigniew Jędrzejewski-Szmek]

On Tue, Feb 14, 2017 at 03:41:52PM +0000, Mat Booth wrote:
>  Ticket is still has "need info" status  -- if the info is
> provided, the
>  status should be set to "discuss at next meeting" 
Indeed. I changed the status. (I feels a bit presumptuous to tell the
FPC when exactly they have to discuss something, but if it's that's 
what it takes, then OK.)

Zbyszek

[Jason L Tibbitts III]

> >>>> "ZJ" == Zbigniew Jędrzejewski-Szmek
> <zbyszek(a)in.waw.pl&gt; writes: 
ZJ> Indeed. I changed the status. (I feels a bit presumptuous to tell
ZJ> the FPC when exactly they have to discuss something, but if it's
ZJ> that's what it takes, then OK.)

trac unfortunately doesn't have any facility for auto-moving things back
from the needinfo status like bugzilla does, so occasionally things get
missed.  Trac will very soon be irrelevant for FPC-related things
anyway, but of course pagure doesn't really even have a concept of
"needinfo" at this point so we'll need to do something else.

 - J<

[Jason L Tibbitts III]

> >>>> "AW" == Adam Williamson
> <adamwill(a)fedoraproject.org&gt; writes: 
AW> Hi folks! So I got bitten again today by the situation where the
AW> primary contact for a given package considers the 'canonical' source
AW> for the spec file to be some external SCM, and finds it a problem
AW> when someone (e.g. a provenpackager like me...) changes the package
AW> directly in dist-git.

But, uh, how does rel-eng do it?  Or do these maintainers all yell at
rel-eng after every mass rebuild?

There are packages where people might want to request communication
before people do work on the package.  Those packages might be
especially complex or have special bootstrapping requirements or a
delicate dependency chain.  That I can understand.  The spec file in
Fedora git not being "canonical", though, is simply not a valid reason.
For Fedora's workflow, which involves a community of package maintainers
who expect to be able to make use of the Fedora's infrastructure and
tools, there can be no possible alternate canonical location for the
spec.

Someone who really, really wants to maintain a package in such a fashion
must be prepared to merge back every commit made to the Fedora package.

 - J<

[Adam Williamson]

On Tue, 2017-02-14 at 18:40 -0600, Jason L Tibbitts III wrote:
>  > > > > > "AW" == Adam Williamson
> <adamwill(a)fedoraproject.org&gt; writes:
>  
>  AW> Hi folks! So I got bitten again today by the situation where the
>  AW> primary contact for a given package considers the 'canonical' source
>  AW> for the spec file to be some external SCM, and finds it a problem
>  AW> when someone (e.g. a provenpackager like me...) changes the package
>  AW> directly in dist-git.
>  
>  But, uh, how does rel-eng do it?  Or do these maintainers all yell at
>  rel-eng after every mass rebuild? 
Yes, more or less. The good ones notice that the mass rebuild happened
and merge the change back. The bad ones just push back over the top of
the mass rebuild and mess stuff up.

>  There are packages where people might want to request communication
>  before people do work on the package.  Those packages might be
>  especially complex or have special bootstrapping requirements or a
>  delicate dependency chain.  That I can understand.  The spec file in
>  Fedora git not being "canonical", though, is simply not a valid reason.
>  For Fedora's workflow, which involves a community of package maintainers
>  who expect to be able to make use of the Fedora's infrastructure and
>  tools, there can be no possible alternate canonical location for the
>  spec. 
This is more or less my opinion, but there are those who don't agree
and have some kind of reason (such as using the same spec file to do
package builds in some other place, or wanting to maintain the spec
file for a tool alongside its source so they can easily do test package
builds as part of the tool's unit tests, etc.)
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net