Re: [Ltib] LFS version

[Stuart Hughes]

Hi Jürgen,

[trimmed back for clarity]

On 04/03/11 08:21, Jürgen Lambrecht wrote:
> On 03/03/2011 09:33 PM, Stuart Hughes wrote:
>> Hi Jurgen,
>>

> If I understand it well (I'm just learning ltib), to work on the source
> code, you must provide a linux tarball, keep the sources in
> ltib/rpm/BUILD/linux, apply your changes, run './ltib -p kernel -m
> patchmerge'. This automatically creates a patch file and updates the
> .spec file.
> But, if I want to commit my patches back to linux, I must use GIT, and
> apply the same changes to the git tree, create a patch...
> Am I correct?
> 

That's mostly correct.  When you start creating a package, you don't
need to provide a tarball if you comment out the Sources tag in the
.spec file.

A better approach for an existing package is to import it.  Download the
srpm (I have had success with fc9) and then run:

$ ltib -m addsrpms _path_to_.src.rpm

This will convert the content and do the LTIB plumbing for the config
system.  You will normally have to hand-edit to fixup (simplify) the
dist/lfs-5.1/_package_/_package_.spec that is converted/created.

Regarding submitting back, here's the crucial thing to remember with LTIB:

* The LTIB tool application is completely separate from the content that
it references.

What this means, is for the changes to the LTIB source code, you can
send a patch to the list by doing a 'cvs diff -u > my_changes.patch'

As far as the content goes, that needs to be uploaded to the GPP by
someone with write access.  Likely some of the content is already there,
in which case you can just send the patches to the list and I will
upload them.  If after a while or working with LTIB it's clear that you
understand the process, you may be granted CVS write and GPP upload
permissions.

BTW: to find out if your referenced content is on the GPP you can run this:

$ ./ltib -p _package_name --dltest

This will show which ones are there and which are not.  Thus you know
which ones need uploading.

Regards, Stuart