From Fedora Project Wiki

< FWN‎ | Beats

Revision as of 19:04, 5 April 2009 by Ush (talk | contribs) (170 devel pass1)


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

Contributing Writer: Oisin Feeley

Noarch with pkconfig Files

PeterRobinson asked[1] for help building his <package>gupnp-vala</package> package as noarch. The complication was that it contained a pkgconfig file.

Several helpful responses, such as MichaelSchwendt's[2] suggested installing pkgconfig files into /usr/share/pkgconfig instead of one of the /usr/lib directories. ToshioKuratomi thoughtCite error: Invalid <ref> tag; refs with no name must have content

VilleSkyttä ran[3] the rpmlint check and confirmed that it warned exactly of this misuse of a libdir macro.

In response to a subsidiary question JesseKeating explained[4] that the noarch packages merely appeared to be present in each of the different architecture trees because they were hardlinked.

Fedora and OpenSolaris Dualboot Issue Solved

After AhmedKamal reported[5] that a ZFS formatted partition seemed to be causing a Fedora 11 Beta installation failure there was a quick response. EricSandeen noted[6] that a patch had already been produced[7] by DaveLehman to merely log the problem instead of raising an error. The bugzilla entry suggested[8] that the root problem was due to udev failing to recognize ZFS properly.

fallocate(2) Preferred Glibc Interface for Preallocation ?

JamesRalson noted[9] the adoption of the ext4 filesystem in Fedora 11 and suggested that in order to use its preallocation features more efficiently it would be useful to patch applications. This could help avoid the current "double write" penalty currently incurred[10] by preallocation in which the reserved space is first filled with nulls. James wondered whether there was a better interface to do this than glibc's posix_fallocate() which first attempts the allocation and then falls "[...] back to writing nulls to fill up the requested range if fallocate() fails."

EricSandeen suggested[11] using fallocate(2) which is present in the glibc version in rawhide and provided a test program to investigate how well this would work.

Rawhide Report Glitches Resolved

After a few "Rawhide Reports" were missed AlexLancaster askedCite error: Invalid <ref> tag; refs with no name must have content what was going on. JoshBoyer answered[12] that pungi for i386 was failing.

Rawhide Reports resumed[13] on 2009-04-04.

XULRunner Committable by non-Provenpackagers

The summary of the 2009-04-03 FESCo meeting indicated[14] that "Firefox/Thunderbird/XULRunner" are open for commits by those who do not have "provenpackager" status. Also discussed and declined for such changes were: popt; initscripts; ethtool; lvm-related packages; and hwdata.

JonStanley also noted[15] that he was going to shoulder the burden of providing his excellent summaries of FESCo meetings.

Provenpackager Policies

Also discussed in the 2009-04-04 FESCo meeting were several requests for "provenpackager" and "sponsor" status. This followed[16] on the heels of work done by PatriceDumas to codify some meanings and processes around "provenpackagers".

A general concern was expressedCite error: Invalid <ref> tag; refs with no name must have content in the IRC meeting that the ability of a provenpackager to modify others' packages should not be used lightly. DavidWoodhouse warned that "provenpackagers who commit to other packages without even _trying_ to coordinate with the owner should expect censure" and JonStanley posted a helpful link[17] to a wiki entry on "Who is allowed to modify which packages".

Python3K Planning

ToshioKuratomi reported[18] on a PyCon[19] talk on Python 3 incompatibility which he had attended. LennartRegebro's "Python 3 Compatibility"[20] talk stimulated Toshio to consider how to port older python code to python-2.6's py3 compatiblity layer.

When JochenSchmitt suggested a compatibility package TomCallaway replied[21] that this would just be a crutch that perpetuated upstream projects unwillingness to move to Python 3. Tom preferred that Fedora developers would "[...] help port such applications to Python 3, and do so in a way that they detect the version of python at runtime and set defines appropriately. That way, we can have applications ready for Python3 before we actually make the switch."

There seemed[22] to be rough agreement between ToshioKuratomi and JamesAntill that some way of allowing python3 modules and an interpreter in parallel to python-2 would be necessary.

IgnacioVazquezAbrams linked[23] to video of all the PyCon 2009 sessions.