Unversioned and >/=/>= obsoletes

[Igor Gnatenko]

All guidelines mandate the use of </<= Obsoletes, but unfortunately we
have some number of packages (179 source rpms -> 292 binary rpms) with
unversioned Obsoletes or with >/=/>= Obsoletes.

It is causing problems with upgrade (if package is getting re-added)
or with 3rd-party repositories. Older package is obsoleting new
package.

Problem categories (in following text by "never" I mean latest N-2 releases):

* Package/SubPackage was never built in Fedora
Package "python" has "Obsoletes: python2" which was never built ->
drop Obsoletes
SubPackage "qpid-proton-c" of "qpid-proton" has "Obsoletes:
qpid-proton" which was not the package for long time -> drop Obsoletes

* Package replacement
Package "storaged" has "Obsoletes: udisks2" -> take latest version
from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
storaged is not simple use-case as it replaces udisks2, but latter is
still not retired.

* "=" Obsoletes
"rubygem-vte" has "Obsoletes: ruby-vte = 3.0.9-1.fc26" (probably
it's
macro in spec) which seems really weird as it will not obsolete
F24/F25 with such version

* Obsoletes by Provides
It doesn't work to prevent undefined behavior. Imagine you have
installed "A" and "B", both providing "C". Package
"D" has "Obsoletes:
C", it should not remove "A" and "B".
** %{?_isa}
"glibc-headers" has "Obsoletes: glibc-headers(i686)". %{?_isa} is
just
text, it's not part of architecture or something else.
** Other provides
"rubygem-http_connection" has "Obsoletes:
rubygem(right_http_connection)". Latter is virtual provides.

* Weird obsoletes (broken)
"krb5-server" has "Obsoletes: krb5-server-1.14.3-8.fc26.i686".
Basically it will not obsolete anything because it's threated as
package name (and we definitely don't have such package name).

* >/>= Obsoletes
"vdsm" has "Obsoletes: vdsm-infra >= 4.16.0". It's almost same
as
unversioned Obsoletes. So it must not be used.

Table of affected packages/maintainers:
https://ignatenkobrain.fedorapeople.org/broken-obsoletes/2016-09-02/broke...
-- 
-Igor Gnatenko

[Stephen Gallagher]

On 09/02/2016 07:14 AM, Igor Gnatenko wrote:
>  All guidelines mandate the use of </<= Obsoletes, but
> unfortunately we
>  have some number of packages (179 source rpms -> 292 binary rpms) with
>  unversioned Obsoletes or with >/=/>= Obsoletes.
>  
>  It is causing problems with upgrade (if package is getting re-added)
>  or with 3rd-party repositories. Older package is obsoleting new
>  package.
>  
>  Problem categories (in following text by "never" I mean latest N-2 releases):
>  
>  * Package/SubPackage was never built in Fedora
>  Package "python" has "Obsoletes: python2" which was never built
> ->
>  drop Obsoletes
>  SubPackage "qpid-proton-c" of "qpid-proton" has "Obsoletes:
>  qpid-proton" which was not the package for long time -> drop Obsoletes
>  
>  * Package replacement
>  Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
>  from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
>  storaged is not simple use-case as it replaces udisks2, but latter is
>  still not retired.
>   
Right, the official plan on this is to retire udisks2 immediately prior to Beta
Freeze[1]

That said, the Obsoletes *should* be versioned. That way if we opted down the
line to re-add it (or change storaged's name, etc.) it could be done.


[1]
https://lists.fedoraproject.org/archives/list/[e-mail].o...


>  * Weird obsoletes (broken)
>  "krb5-server" has "Obsoletes: krb5-server-1.14.3-8.fc26.i686".
>  Basically it will not obsolete anything because it's threated as
>  package name (and we definitely don't have such package name).
>   
This definitely looks odd... Robbie?

[Richard W.M. Jones]

On Fri, Sep 02, 2016 at 01:14:13PM +0200, Igor Gnatenko wrote:
>  Table of affected packages/maintainers:
>  https://ignatenkobrain.fedorapeople.org/broken-obsoletes/2016-09-02/broke...
I fixed open-vm-tools, only in dist-git.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW

[Michael Catanzaro]

On Fri, 2016-09-02 at 13:14 +0200, Igor Gnatenko wrote:
>  * Package replacement
>  Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
>  from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
>  storaged is not simple use-case as it replaces udisks2, but latter is
>  still not retired. 
Everyone OK if I retire it in rawhide now? I think if we were going to
have any issue that would cause us to switch back to udisks, it would
have manifested by now.

[Stephen Gallagher]

On 09/02/2016 07:55 AM, Michael Catanzaro wrote:
>  On Fri, 2016-09-02 at 13:14 +0200, Igor Gnatenko wrote:
> > * Package replacement
> > Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
> > from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
> > storaged is not simple use-case as it replaces udisks2, but latter is
> > still not retired.
>  
>  Everyone OK if I retire it in rawhide now? I think if we were going to
>  have any issue that would cause us to switch back to udisks, it would
>  have manifested by now. 
Makes sense to me. Do we want to do F25 at the same time or wait until closer to
Beta Freeze (2016-09-27)?

[Kalev Lember]

On 09/02/2016 02:36 PM, Stephen Gallagher wrote:
>  On 09/02/2016 07:55 AM, Michael Catanzaro wrote:
> > On Fri, 2016-09-02 at 13:14 +0200, Igor Gnatenko wrote:
> >> * Package replacement
> >> Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
> >> from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
> >> storaged is not simple use-case as it replaces udisks2, but latter is
> >> still not retired.
> >
> > Everyone OK if I retire it in rawhide now? I think if we were going to
> > have any issue that would cause us to switch back to udisks, it would
> > have manifested by now.
>  
>  Makes sense to me. Do we want to do F25 at the same time or wait until closer to
>  Beta Freeze (2016-09-27)? 
Please do it now if you plan to do it for F25, so that if there's any
fallout we get to find out about it now and have time to fix things. If
it lands close to a freeze there probably won't be any time to fix
things up before the freeze hits us.

-- 
Kalev

[Stephen Gallagher]

On 09/02/2016 08:44 AM, Kalev Lember wrote:
>  On 09/02/2016 02:36 PM, Stephen Gallagher wrote:
> > On 09/02/2016 07:55 AM, Michael Catanzaro wrote:
> >> On Fri, 2016-09-02 at 13:14 +0200, Igor Gnatenko wrote:
> >>> * Package replacement
> >>> Package "storaged" has "Obsoletes: udisks2" -> take
> latest version
> >>> from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
> >>> storaged is not simple use-case as it replaces udisks2, but latter is
> >>> still not retired.
> >>
> >> Everyone OK if I retire it in rawhide now? I think if we were going to
> >> have any issue that would cause us to switch back to udisks, it would
> >> have manifested by now.
> >
> > Makes sense to me. Do we want to do F25 at the same time or wait until closer to
> > Beta Freeze (2016-09-27)?
>  
>  Please do it now if you plan to do it for F25, so that if there's any
>  fallout we get to find out about it now and have time to fix things. If
>  it lands close to a freeze there probably won't be any time to fix
>  things up before the freeze hits us.
>   
Well, if there was fallout during the Beta period, there would still be Final to
revert it, but in general I agree: let's do it sooner rather than later, while
we have more time to react.

[Michael Catanzaro]

On Fri, 2016-09-02 at 08:47 -0400, Stephen Gallagher wrote:
>  Well, if there was fallout during the Beta period, there would still
>  be Final to
>  revert it, but in general I agree: let's do it sooner rather than
>  later, while
>  we have more time to react. 
OK, done. Hopefully nothing breaks.

[Michael Catanzaro]

On Fri, 2016-09-02 at 08:36 -0400, Stephen Gallagher wrote:
>  Makes sense to me. Do we want to do F25 at the same time or wait
>  until closer to
>  Beta Freeze (2016-09-27)? 
I just did both rawhide and F25.

[Kevin Kofler]

Michael Catanzaro wrote:
>  Everyone OK if I retire it in rawhide now? I think if we were going
> to
>  have any issue that would cause us to switch back to udisks, it would
>  have manifested by now. 
Actually, there is this one:
http://www.spinics.net/linux/fedora/fedora-kde/msg18000.html

        Kevin Kofler

[Michael Schwendt]

On Fri, 2 Sep 2016 13:14:13 +0200, Igor Gnatenko wrote:

>  All guidelines mandate the use of </<= Obsoletes, but
> unfortunately we
>  have some number of packages (179 source rpms -> 292 binary rpms) with
>  unversioned Obsoletes or with >/=/>= Obsoletes.
>  
>  It is causing problems with upgrade (if package is getting re-added)
>  or with 3rd-party repositories. Older package is obsoleting new
>  package. 
Good luck with trying to get some packagers to fix such issues!
I appreciate the effort as I've reported similar things many times before,
but some packagers just don't respond in bugzilla or overwrite changes
applied to git after waiting months for a reply.

[Igor Gnatenko]

On Fri, Sep 2, 2016 at 3:34 PM, Michael Schwendt <mschwendt(a)gmail.com&gt; wrote:
>  On Fri, 2 Sep 2016 13:14:13 +0200, Igor Gnatenko wrote:
>
> > All guidelines mandate the use of </<= Obsoletes, but unfortunately we
> > have some number of packages (179 source rpms -> 292 binary rpms) with
> > unversioned Obsoletes or with >/=/>= Obsoletes.
> >
> > It is causing problems with upgrade (if package is getting re-added)
> > or with 3rd-party repositories. Older package is obsoleting new
> > package.
>
>  Good luck with trying to get some packagers to fix such issues!
>  I appreciate the effort as I've reported similar things many times before,
>  but some packagers just don't respond in bugzilla or overwrite changes
>  applied to git after waiting months for a reply. 
Isn't this is a guidelines, so
if packager ignores them - he should be punished?
>  --
>  devel mailing list
>  devel(a)lists.fedoraproject.org
>  https://lists.fedoraproject.org/admin/lists/[e-mail] 


-- 
-Igor Gnatenko

[Parag Nemade]

On Fri, Sep 2, 2016 at 7:21 PM, Igor Gnatenko <ignatenko(a)redhat.com&gt; wrote:
>  On Fri, Sep 2, 2016 at 3:34 PM, Michael Schwendt
> <mschwendt(a)gmail.com&gt; wrote:
> > On Fri, 2 Sep 2016 13:14:13 +0200, Igor Gnatenko wrote:
> >
> >> All guidelines mandate the use of </<= Obsoletes, but unfortunately we
> >> have some number of packages (179 source rpms -> 292 binary rpms) with
> >> unversioned Obsoletes or with >/=/>= Obsoletes.
> >>
> >> It is causing problems with upgrade (if package is getting re-added)
> >> or with 3rd-party repositories. Older package is obsoleting new
> >> package.
> >
> > Good luck with trying to get some packagers to fix such issues!
> > I appreciate the effort as I've reported similar things many times before,
> > but some packagers just don't respond in bugzilla or overwrite changes
> > applied to git after waiting months for a reply.
>  Isn't this is a guidelines, so if packager ignores them - he should be punished?
No, packager's sponsor will be contacted and he will guide the
packager for the correct usage of packaging guidelines.

Regards,
Parag

[Josh Boyer]

On Fri, Sep 2, 2016 at 9:51 AM, Igor Gnatenko <ignatenko(a)redhat.com&gt; wrote:
>  On Fri, Sep 2, 2016 at 3:34 PM, Michael Schwendt
> <mschwendt(a)gmail.com&gt; wrote:
> > On Fri, 2 Sep 2016 13:14:13 +0200, Igor Gnatenko wrote:
> >
> >> All guidelines mandate the use of </<= Obsoletes, but unfortunately we
> >> have some number of packages (179 source rpms -> 292 binary rpms) with
> >> unversioned Obsoletes or with >/=/>= Obsoletes.
> >>
> >> It is causing problems with upgrade (if package is getting re-added)
> >> or with 3rd-party repositories. Older package is obsoleting new
> >> package.
> >
> > Good luck with trying to get some packagers to fix such issues!
> > I appreciate the effort as I've reported similar things many times before,
> > but some packagers just don't respond in bugzilla or overwrite changes
> > applied to git after waiting months for a reply.
>  Isn't this is a guidelines, so if packager ignores them - he should be punished?
We have no recourse for punishment.  Frankly, that's not a great plan
anyway.  We should focus on collaboration and education, not punitive
actions.

I would rather focus on fixing the packages.  If the primary contact
can't or won't do it, then a provenpackager should be able to fix it
instead.  If there's a persistent issue with reverts of those kinds of
changes or something, we can figure it out later.

josh

[Charalampos Stratakis]

There is the Non-responsive Maintainer Policy [0] however it takes way too long, and in
the end you will get ownership of the package, so if you want to do a minor fix to a
package (and the maintainer is not responding), you will either need to wait 3+ weeks and
then get the package and do it yourself, or bug provenpackagers to do the change for you.

Then of course people will either complain that they didn't see the bugzilla's for
x,y reasons or that they didn't want the change from a proven packager.

Not sure if there is a perfect solution that would satisfy everyone for this case.

[0] https://fedoraproject.org/wiki/Policy_for_nonresponsive_package_maintainers

Charalampos Stratakis
Associate Software Engineer
Python Maintenance Team, Red Hat


----- Original Message -----
From: "Igor Gnatenko" <ignatenko(a)redhat.com&gt;
To: "Development discussions related to Fedora"
<devel(a)lists.fedoraproject.org&gt;
Sent: Friday, September 2, 2016 3:51:54 PM
Subject: Re: Unversioned and >/=/>= obsoletes

On Fri, Sep 2, 2016 at 3:34 PM, Michael Schwendt <mschwendt(a)gmail.com&gt; wrote:
>  On Fri, 2 Sep 2016 13:14:13 +0200, Igor Gnatenko wrote:
>
> > All guidelines mandate the use of </<= Obsoletes, but unfortunately we
> > have some number of packages (179 source rpms -> 292 binary rpms) with
> > unversioned Obsoletes or with >/=/>= Obsoletes.
> >
> > It is causing problems with upgrade (if package is getting re-added)
> > or with 3rd-party repositories. Older package is obsoleting new
> > package.
>
>  Good luck with trying to get some packagers to fix such issues!
>  I appreciate the effort as I've reported similar things many times before,
>  but some packagers just don't respond in bugzilla or overwrite changes
>  applied to git after waiting months for a reply. 
Isn't this is a guidelines, so
if packager ignores them - he should be punished?
>  --
>  devel mailing list
>  devel(a)lists.fedoraproject.org
>  https://lists.fedoraproject.org/admin/lists/[e-mail] 


-- 
-Igor Gnatenko
--
devel mailing list
devel(a)lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/[e-mail]

[Tomáš Smetana]

On Fri, 2 Sep 2016 13:14:13 +0200
Igor Gnatenko <ignatenko(a)redhat.com&gt; wrote:

>  * Package replacement
>  Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
>  from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
>  storaged is not simple use-case as it replaces udisks2, but latter is
>  still not retired. 
Followed your advice, should be fixed soon in Rawhide and F25.

Thanks and regards,
-- 
Tomáš Smetana
Platform Engineering, Red Hat

[Charalampos Stratakis]

Fixed in python and python-docs

Charalampos Stratakis
Associate Software Engineer
Python Maintenance Team, Red Hat


----- Original Message -----
From: "Igor Gnatenko" <ignatenko(a)redhat.com&gt;
To: "Development discussions related to Fedora"
<devel(a)lists.fedoraproject.org&gt;, devel-announce(a)lists.fedoraproject.org
Sent: Friday, September 2, 2016 1:14:13 PM
Subject: Unversioned and >/=/>= obsoletes

All guidelines mandate the use of </<= Obsoletes, but unfortunately we
have some number of packages (179 source rpms -> 292 binary rpms) with
unversioned Obsoletes or with >/=/>= Obsoletes.

It is causing problems with upgrade (if package is getting re-added)
or with 3rd-party repositories. Older package is obsoleting new
package.

Problem categories (in following text by "never" I mean latest N-2 releases):

* Package/SubPackage was never built in Fedora
Package "python" has "Obsoletes: python2" which was never built ->
drop Obsoletes
SubPackage "qpid-proton-c" of "qpid-proton" has "Obsoletes:
qpid-proton" which was not the package for long time -> drop Obsoletes

* Package replacement
Package "storaged" has "Obsoletes: udisks2" -> take latest version
from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
storaged is not simple use-case as it replaces udisks2, but latter is
still not retired.

* "=" Obsoletes
"rubygem-vte" has "Obsoletes: ruby-vte = 3.0.9-1.fc26" (probably
it's
macro in spec) which seems really weird as it will not obsolete
F24/F25 with such version

* Obsoletes by Provides
It doesn't work to prevent undefined behavior. Imagine you have
installed "A" and "B", both providing "C". Package
"D" has "Obsoletes:
C", it should not remove "A" and "B".
** %{?_isa}
"glibc-headers" has "Obsoletes: glibc-headers(i686)". %{?_isa} is
just
text, it's not part of architecture or something else.
** Other provides
"rubygem-http_connection" has "Obsoletes:
rubygem(right_http_connection)". Latter is virtual provides.

* Weird obsoletes (broken)
"krb5-server" has "Obsoletes: krb5-server-1.14.3-8.fc26.i686".
Basically it will not obsolete anything because it's threated as
package name (and we definitely don't have such package name).

* >/>= Obsoletes
"vdsm" has "Obsoletes: vdsm-infra >= 4.16.0". It's almost same
as
unversioned Obsoletes. So it must not be used.

Table of affected packages/maintainers:
https://ignatenkobrain.fedorapeople.org/broken-obsoletes/2016-09-02/broke...
-- 
-Igor Gnatenko
--
devel mailing list
devel(a)lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/[e-mail]

[Igor Gnatenko]

I'm going to do mass bug filling for those packages which are still
not fixed. Are there some scripts to do that or I have to write my
own?

Unfixed packages:
https://ignatenkobrain.fedorapeople.org/broken-obsoletes/latest/broken-ob...

On Fri, Sep 2, 2016 at 1:14 PM, Igor Gnatenko <ignatenko(a)redhat.com&gt; wrote:
>  All guidelines mandate the use of </<= Obsoletes, but
> unfortunately we
>  have some number of packages (179 source rpms -> 292 binary rpms) with
>  unversioned Obsoletes or with >/=/>= Obsoletes.
>
>  It is causing problems with upgrade (if package is getting re-added)
>  or with 3rd-party repositories. Older package is obsoleting new
>  package.
>
>  Problem categories (in following text by "never" I mean latest N-2 releases):
>
>  * Package/SubPackage was never built in Fedora
>  Package "python" has "Obsoletes: python2" which was never built
> ->
>  drop Obsoletes
>  SubPackage "qpid-proton-c" of "qpid-proton" has "Obsoletes:
>  qpid-proton" which was not the package for long time -> drop Obsoletes
>
>  * Package replacement
>  Package "storaged" has "Obsoletes: udisks2" -> take latest
> version
>  from koji (2.1.7-1) and make Obsoletes versioned: udisks2 < 2.1.7-2
>  storaged is not simple use-case as it replaces udisks2, but latter is
>  still not retired.
>
>  * "=" Obsoletes
>  "rubygem-vte" has "Obsoletes: ruby-vte = 3.0.9-1.fc26" (probably
> it's
>  macro in spec) which seems really weird as it will not obsolete
>  F24/F25 with such version
>
>  * Obsoletes by Provides
>  It doesn't work to prevent undefined behavior. Imagine you have
>  installed "A" and "B", both providing "C". Package
> "D" has "Obsoletes:
>  C", it should not remove "A" and "B".
>  ** %{?_isa}
>  "glibc-headers" has "Obsoletes: glibc-headers(i686)". %{?_isa} is
> just
>  text, it's not part of architecture or something else.
>  ** Other provides
>  "rubygem-http_connection" has "Obsoletes:
>  rubygem(right_http_connection)". Latter is virtual provides.
>
>  * Weird obsoletes (broken)
>  "krb5-server" has "Obsoletes: krb5-server-1.14.3-8.fc26.i686".
>  Basically it will not obsolete anything because it's threated as
>  package name (and we definitely don't have such package name).
>
>  * >/>= Obsoletes
>  "vdsm" has "Obsoletes: vdsm-infra >= 4.16.0". It's almost same
> as
>  unversioned Obsoletes. So it must not be used.
>
>  Table of affected packages/maintainers:
>  https://ignatenkobrain.fedorapeople.org/broken-obsoletes/2016-09-02/broke...
>  --
>  -Igor Gnatenko 


-- 
-Igor Gnatenko

[Kevin Fenzi]

On Tue, 13 Sep 2016 11:38:24 +0200
Igor Gnatenko <ignatenko(a)redhat.com&gt; wrote:

>  I'm going to do mass bug filling for those packages which are
> still
>  not fixed. Are there some scripts to do that or I have to write my
>  own?
>  
>  Unfixed packages:
>  https://ignatenkobrain.fedorapeople.org/broken-obsoletes/latest/broken-ob...
Can you please take a look at: 

https://fedoraproject.org/wiki/Mass_bug_filing

and address the items there before filing?

Happy to help review a proposed bug text/announcement.

kevin