F-10 joyride vs 8.2 - getting fixes upstream.
pbrobinson at gmail.com
Tue Nov 18 04:34:16 EST 2008
>> also notice that the "Size delta" on the 8.2 vs joyride build is
>> reporting 0meg difference which is clearly rubbish :-) Can someone fix
>> it for me please. The URL I'm referring to is
> This one works:
>> I should get more time on Wednesday to continue through the list. I'll
>> be adding any bugs I file against the tracker bug which can be found
>> here https://bugzilla.redhat.com/show_bug.cgi?id=462625
> I am planning on working on repeating a few forks that I made during
> the F9 cycle to fix bugs and reduce dependencies. All of these changes
> are not easy to fix upstream or in F10.
I agree with Jeremy on this one. And I don't necessarily agree that in
all cases that they aren't easy to fix upstream. I've been through
most of the forks from Fedora 9 to OLPC 8.2 and the vast majority of
them are actually pretty easy fixes. Some of them are things like
splitting functionality out into sub packages. Some I have already
filed bugs against.
> For example, totem-pl-parser pulling in evolution-data-server and
> libgnome and ...
> There is already a (closed) bug about this blocking the one you
> mentioned: #456113
> Is that sufficient for your tracking, or should I reopen it, or create
> a new one?
If we do need to fork I personally would like it done on existing bugs
because it allows for easy history tracking.
> I guess I will have to create another branch request to carry this
> out.. should I make that bug block FedoraOLPCDelta?
Any bugs should block the OLPC Delta.
Looking at the increase in size from 8.2 to the current 2550 joyride I
suspect the 90 odd meg of bloat falls into a couple of categories.
Perl being the first which probably accounts for nearly half of that,
then there's new core OS package additions like plymouth, and then the
dependency chains that have bloated again like libgnome, e-d-s etc.
Alot of the later should be quite simple to fix, especially as
libgnome etc are starting to ramp up in their disappearing act.
The perl dependency should be fairly easy to get rid of, I suspect
that there's only 2-3 packages such as abiword and ghostscript that
are pulling it in, and I think I have bugs for both of those. That
will cut down the size delta considerably and give us some breathing
space to deal with the rest of them. Gstreamer is I know is pulling in
some new dependencies some of which might be wanted (wav support) and
others not (cd ripping support) which should be able to be separated
out into a sub package.
If you can give me a bit of time to get some traction on this I should
in the next couple of days be able to get a better idea on where we
sit. Then an email to the fedora-olpc list for some help and I think
the vast majority of the problems will be gone.
The likes of totem-pl-parser might take a bit longer so we might have
to in the short term fork but the less we have to do now, the less
pain we have in the future. And by using mainline packages it also
allows us to monitor closely through joyride any of the dependency
bloat and jump on mainline before it becomes an issue.
More information about the Devel