From Fedora Project Wiki

< FWN‎ | Beats

Line 66: Line 66:
[[RichardJones|Richard W.M. Jones]] announced<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01055.html</ref> the availability of <code>CrossReport</code>, a tool to evaluate the ease with which applications can be ported to Windows using the <code>MinGW</code> libraries.
[[RichardJones|Richard W.M. Jones]] announced<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01055.html</ref> the availability of <code>CrossReport</code>, a tool to evaluate the ease with which applications can be ported to Windows using the <code>MinGW</code> libraries.


After some issue with platform dependency were reported by [[MichaelCronenworth|Michael Cronenworth]] were sorted<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01074.html</ref> out it seemed<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01076.html</ref> the tool is ready for use.
After some issues with platform dependency were reported by [[MichaelCronenworth|Michael Cronenworth]] and quickly sorted<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01074.html</ref> out it seemed<ref>http://www.redhat.com/archives/fedora-devel-list/2009-February/msg01076.html</ref> the tool is ready for use.


<references/>
<references/>

Revision as of 15:42, 16 February 2009

Developments

In this section the people, personalities and debates on the @fedora-devel mailing list are summarized.

Contributing Writer: Oisin Feeley

FLOSS Multimedia Codec Support

Inspired by previous discussions on whether Fedora should distribute FLOSS content[1] Martin Sourada tried[2] to start a discussion about the poor support of FLOSS multimedia. Martin noted: "Out of the combinations of two FLOSS containers (matroska and ogg) and two FLOSS video codecs (dirac and theora) I know only one (ogg + theora) actually works in xine-lib (used by KDE4) which is pathetic." He asked for help in documenting the situation on a wiki page[3].

When Bastien Nocera suggested that the most important thing was to file bugs Martin responded[4] that this was what he was doing after first performing tests.

An information packed sub-thread started[5] by Gregory Maxwell expanded the scope of the tests and started a discussion with Dominik Mierzejewski about the problem of ffmpeg providing sub-optimal implementations of unencumbered codecs. It seems that for reasons of efficiency ffmpeg re-invents the wheel from scratch instead of using and improving upstream implementations. Kevin Kofler also rose[6] to the implied challenge that GStreamer was preferable to xine-lib.

Multiple Packages from One Source ?

A question about how to handle the situation where a single source could be compiled with alternate databases was posted[1] by Steven Moix. The motion video motion detector software can be compiled to use either MySQL or Postgres. Steven explained that the problem was that "[...]you can't divide it into sub-packages, at the end it generates one big binary file [...]" and wondered should he just choose the database he preferred or propose two packages.

Manuel Wolfshant expressed[2] what appeared to be the common wisdom: "personally I would compile twice, once enabling mysql and another time pgsql, and create 2 packages. each package would install a "motion-dbname" binary, and a symlink would allow access via the well known name "motion". Using alternatives would allow a switch between the two."

Although it was admitted that David Woodhouse's suggestion[3] to make the program use loadable plugins was the ideal Tom Lane thought[4] that was "[...] a bit above and beyond what a packager should do. If he's also an upstream developer, then he should undertake that addition with his developer hat on; but it's *well* beyond the size of patch that a Fedora package should be carrying."

The ability to specify alternate requires (similar to those used in the .deb package format[5]) was discussed[6] by Richard W.M. Jones and Tom Lane and dismissed as impractical in this case anyway due to variations in SQL.

Take a Peek at the Fedora 11 Release Notes

Fresh from his work on the RHEL-5.3 Release Notes Ryan Lerch apprised[1] the list of the latest changes to the Fedora 11 Release Notes. Ryan sought early feedback and changes to documentation beats in order to give the community an early preview of the release notes.

Initial feedback from Thorsten Leemhuis and Kevin Kofler and others indicated that the use of fixed-width instead of liquid layout was disliked by some people and loved[2] by others.

Ryan also provided[3] an rpm of this Release Notes mockup.

Heads Up: Noarch Subpackages Landing Soon

Florian Festi warned[1] that the ability to produce noarch subpackages will soon be available in Fedora. This brings the benefit of being able to share these packages among different architectures thus reducing disk space and mirror bandwidth.

Although rpm-4.6 supports noarch fully there are still some fixes to make to koji before the Fedora buildsystem can cope with noarch subpackages. Florian suggested that those who wanted to could experiment in mock with rpmdiff to compare the results across different architectures. He assured readers that there were no plans to force packagers to use this feature and invited anyone interested in developing the use of noarch in future release to a discussion.

Florian later warned[2] that one potential negative outcome of using such sub-packages would be a proliferation of packages and consequent bloating of metadata which might affect yum.

VilleSkyttä suggested[3] that it would be wise to make sure that use of BuildRequire: rpm-build >= 4.6.0 was enforced in order to ensure that earlier versions of rpmbuild did not produce noarch versions of the main package and other potential subpackages. Florian's response recognized[4] the problem but deprecated the use of BuildRequires to such an extent. One possible alternative which he proposed was to have Panu Matilainen "[...] backport a check that will make noarch packages (both regular and noarch) fail to build if they contain binaries [and ensure that this] additional check will be in place before koji will be updated[.]" This latter proposal stimulated a good deal of interest from Ralf Corsepius and Richard W.M. Jones as they were both concerned with cross-architecture issues. The definition of a "binary" seemed to be one unclear point.

In a later thread Florian updated[5] a list of packages which could be easily turned into noarch subpackages.

Mass Rebuild Coming Soon

Jesse Keating drew attention to "[...] a perfect storm brewing for Fedora 11" due to the need to rebuild with GCC-4.4 (see FWN#161[1], the use of i586 as the default supported architecture (see FWN#162[2]) and the support of stronger hashes (last paragraph of FWN#107[3]).

Apparently the time-constraints led to a desire to start the rebuild as soon as possible without giving maintainers an explicit window in which to do their own builds. Jesse preferred to give maintainers an ability to opt-out and sought suggestions on how this could be achieved.

Jesse suggested that interested parties should either reply to the thread and/or participate in the 2009-02-16 IRC meeting in #fedora-meeting at 1800UTC.

New Tool Measures Ease of Cross-compiling to Windows

Richard W.M. Jones announced[1] the availability of CrossReport, a tool to evaluate the ease with which applications can be ported to Windows using the MinGW libraries.

After some issues with platform dependency were reported by Michael Cronenworth and quickly sorted[2] out it seemed[3] the tool is ready for use.