Bug 225635

GOOD
====
MUST: Naming Guidelines followed
MUST: Packaging Guidelines generally followed (except SMP flags below)
MUST: spec filename is fine
MUST: License of the library itself is open source (LGPL/MPL)
MUST: License filed matches actual licenses
MUST: spec file legible
MUST: sources match upstream (md5sum used)
MUST: RPM build for i386
MUST: no ExcludeArch
MUST: build dependencies fine
MUST: no locales
MUST: ldconfig called fine
MUST: not relocatable
MUST: no duplicate files
MUST: file permissions fine (except for -debuginfo, mentioned below)
MUST: %clean section fine
MUST: macro use consistent
MUST: package contains code
MUST: large documentation files in -devel instead of -doc (acceptable)
MUST: %doc does not affect run-time
MUST: header files in -devel
MUST: -devel requries pkgconfig
MUST: *.so in -devel
MUST: -devel has fully versioned dependency
MUST: *.la removed explicitly
MUST: not GUI
MUST: doesn't own others' files or dirs
SHOULD: no scriptlets
SHOULD: no subpackage other than -devel
SHOULD: pkgconfig file in -devel

BAD
===
MUST: rpmlint output:

  $ rpmlint cairo-1.3.12-1.i386.rpm 
  E: cairo obsolete-not-provided libpixman
  E: cairo obsolete-not-provided libpixman-devel
  E: cairo obsolete-not-provided libpixman-debuginfo

  (Provides should be added, -devel and -debuginfo should be obsoleted and
  provided by subpackages)

  $ rpmlint cairo-devel-1.3.12-1.i386.rpm 
  W: cairo-devel no-documentation

  (The HTML files provided in the package are definitely documentation and
  should be marked as such.)

  $ rpmlint cairo-debuginfo-1.3.12-1.i386.rpm 
  W: cairo-debuginfo spurious-executable-perm
/usr/src/debug/cairo-1.3.12/src/cairo-scaled-font.c
  W: cairo-debuginfo spurious-executable-perm
/usr/src/debug/cairo-1.3.12/src/cairoint.h

  (These two files are wrongly set to be executables.)

MUST: licensing
 
  1) The license files COPYING-LGPL-2.1 and COPYING-MPL-1.1 are not included
  in the binary RPM.
  
  2) Some auxiliary files refer to GPL (like test/pdiff/pdiff.c),
  and say that one should have received a copy of the license with the file,
  while it is not available in the package itself. Upstream problem, but a
  SHOULD in review list (contact upstream).

  3) Although the COPYING file mentions that all "auxiliary components" (test
  files etc.) are free software and refers to the files' headers, some are
  not, because their headers doesn't mention any such licensing, which makes
  them proprietary. Examples: composite-integer-translate-over-repeat.c,
  composite-integer-translate-source.c, ... in the tests subdirectory.

MUST: US English

  Suggestions: replace "eg." with "e.g." or "for example"; capitalize "cairo"
  to "Cairo" in the -devel package description.

MUST: owning directories

   The -devel subpackage installs files in /usr/share/gtk-doc/html without
   owning the directory or depending on anything that owns it.

MUST: packaging guidelines

   * make should use %{?_smp_mflags} flag.
   
SUGGESTIONS
===========
* Consider shipping some of the not-shipped docs, like PORTING_GUIDE or
  TODO as documentation.

(In reply to comment #1)
> BAD
> ===
> MUST: rpmlint output:
> 
>   $ rpmlint cairo-1.3.12-1.i386.rpm 
>   E: cairo obsolete-not-provided libpixman
>   E: cairo obsolete-not-provided libpixman-devel
>   E: cairo obsolete-not-provided libpixman-debuginfo
>
>   (Provides should be added, -devel and -debuginfo should be obsoleted and
>   provided by subpackages)

This is intentional and not error.  We are not providing libpixman in cairo. 
Hiding it.  Anybody depending on libpixman is doomed.  But the only user was
cairo.  So that's not a problem in reality.
 
>   $ rpmlint cairo-debuginfo-1.3.12-1.i386.rpm 
>   W: cairo-debuginfo spurious-executable-perm
> /usr/src/debug/cairo-1.3.12/src/cairo-scaled-font.c
>   W: cairo-debuginfo spurious-executable-perm
> /usr/src/debug/cairo-1.3.12/src/cairoint.h

Weird.  Not sure howthis is happening.
 
> MUST: US English
> 
>   Suggestions: replace "eg." with "e.g." or "for example"; capitalize "cairo"
>   to "Cairo" in the -devel package description.

Carl advertises for non-capitalized name usage.  Mostly to differentiate from
the lesser known Cairo.

> * Consider shipping some of the not-shipped docs, like PORTING_GUIDE or
>  TODO as documentation.

Neither PORTING_GUIDE nor TODO are useful documentation for users of cairo.

PORTING_GUIDE makes sense in -devel, if it was not so obsolete these days.

Some packages ship TODO, other don't.

(In reply to comment #2)
> This is intentional and not error.  We are not providing libpixman in cairo. 
> Hiding it.  Anybody depending on libpixman is doomed.  But the only user was
> cairo.  So that's not a problem in reality.

What is the intended user experience here?
* What should happen if a user has libpixman and old cairo installed and then
  tries to update his box to a repository that contains the present version of
  cairo?
* What should happen if he only has libpixman and tries to install cairo from
  the same repository?
* What should happen if a user has a new cairo and then tries to install a
  libpixman RPM (using a .rpm file he found somewhere that was copied from
  rawhide once)?

Having answers for these three cases will help finding the best settings for the
Obsoletes lines.

I don't see libpixman in any of FC2, FC3, FC4, FC5, FC6, thus we should be able
to simply drop this obsolete Obsoletes: - Toshio said that we should not worry
about going back further than 2 Fedora releases.
 

> Toshio said that we should not worry
> about going back further than 2 Fedora releases.

Generally yes. But to be on the very safe side, we may wish to go as far back as
FC3 and FC4 (since they were dropped pre-maturely and may have some people who
have not switched yet), but not further. No difference to libpixman of course.

libpixman was just available in rawhide before FC5, when cairo had not reached 1.0.

(In reply to comment #8)
> libpixman was just available in rawhide before FC5, when cairo had not reached
1.0.

OK. If you want to keep the Obsoletes, please provide answers to comment #5 so
we can make sure the best is specified in the spec. If you want to remove them,
feel free.

now you are being a bit too picky.  I'm going to remove the Obsoletes, but here
are the answers anyway:

(In reply to comment #5)
> What is the intended user experience here?
> * What should happen if a user has libpixman and old cairo installed and then
>   tries to update his box to a repository that contains the present version of
>   cairo?

libpixman removed, as is apparently what happens.

> * What should happen if he only has libpixman and tries to install cairo from
>   the same repository?

"Don't care".  Removed.

> * What should happen if a user has a new cairo and then tries to install a
>   libpixman RPM (using a .rpm file he found somewhere that was copied from
>   rawhide once)?

"Don't care".  Conflict.

> Having answers for these three cases will help finding the best settings for the
> Obsoletes lines.

I only can think of two settings: keep them or remove them.  Removing  defeats
the purpose of what it was there for: to clean up an unneeded package.  Keeping
them does just that.  There's nothing special here as the only package that
evern depended on libpixman was cairo.

But as I said, I'll remove them anyway.  (the situation is similar to pango
obsoleting pango-gtkbeta and fribidi-gtkbeta).

FWIW, I've removed the Obsoletes: glib-gtkbeta from glib2.

Behdad, did you want to make any changes here ?

Just wanted to remove the Obsoletes.  I'll go over my merge review bugs next week.

Mostly fixed.  Remaining issues are:

  $ rpmlint cairo-devel-1.3.12-1.i386.rpm 
  W: cairo-devel no-documentation

  (The HTML files provided in the package are definitely documentation and
  should be marked as such.)

MUST: owning directories

   The -devel subpackage installs files in /usr/share/gtk-doc/html without
   owning the directory or depending on anything that owns it.

MUST: packaging guidelines

   * make should use %{?_smp_mflags} flag.


None of the few packages I looked at (pango, gtk+) does any of them, so I'm not
sure about to do them in the right way.  For one thing, we pass
--disable-gtk-doc to configure, so requiring gtk-doc doesn't make much sense.

Stalled review ?

FWIW, I have started to require gtk-doc in -devel packages now, to get over the
directory ownership annoyance.

How should we proceed.  rpmlint is happy about cairo.  Someone can send a patch for the minor issues left?

Mass reassigning all merge reviews to their component. For more details, see this FESCO ticket:

  https://fedorahosted.org/fesco/ticket/1269

If you don't know what merge reviews are about, please see:

  https://fedoraproject.org/wiki/Merge_Reviews

How to handle this bug is left to the discretion of the package maintainer.

This bug appears to have been reported against 'rawhide' during the Fedora 23 development cycle.
Changing version to '23'.

(As we did not run this process for some time, it could affect also pre-Fedora 23 development
cycle bugs. We are very sorry. It will help us with cleanup during Fedora 23 End Of Life. Thank you.)

More information and reason for this action is here:
https://fedoraproject.org/wiki/BugZappers/HouseKeeping/Fedora23

Closing the ticket, I don't think there's anything left to do here.