On Thu, 2020-04-16 at 13:10 +0200, Nicolas Chauvet wrote:
Le mer. 15 avr. 2020 à 13:34, Josh Boyer jwboyer@fedoraproject.org a écrit :
On Wed, Apr 15, 2020 at 5:32 AM Thorsten Leemhuis < fedora@leemhuis.info> wrote:
...
There is one thing I really dislike about the scheme (one it didn't notice when I took a brief look at it weeks ago; sorry): There are no individual patches anymore in dist-git/the srpm and that afaics violates the packaging guidelines. https://docs.fedoraproject.org/en-US/packaging-guidelines/#_applying_patches
The files MUST then be checked into the Fedora Package revision control system […]. Storing the files in this way allows people to use standard tools to visualize the changes between revisions of the files and track additions and removals without a layer of indirection […].
There are other rules in the patch section that afaics are violated. Were those violation discussed and blessed by the Fedora Packaging Committee or FESCo?
I don't think the split out patches "rule" is being violated here. They changed the source tarball to one generated from the git tree and they don't have any patches at the dist-git level at all. Several other packages in Fedora already do this, such as anaconda.
So it means should one single line patch should be changed, you need to re-upload a 100M tarball to the fedora lookaside cache? I understand the benefit to have an upstream source tree as a primary origin for kernel patch development. But what is the reason behind this process in particular ?
It would have been easier to only generate a full fedora diff against the original upstream tree (using a commit-id that can also be generated with gitlab with A PR). Like how I expect it's done with the patch-5.7.0-redhat.patch.
Then keep using the original upstream tarballs. (linux-M.tar.xz and patch-M-m.xz) I'm sure that in some cases, (minor patch update) there is even not have to change the fedora diff.
Yes, it's a bit inefficient. I brought this up on the infrastructure list ([RFC] Optionally using git repositories instead of the lookaside cache), but people didn't seem concerned about the storage requirement.
I am not particularly inspired to develop a bunch of scripts to work around the fact that we're not using git to store source code. If someone else wants to fix up the kernel build scripts I'm fine with that, although it seems to me that it's fixing the problem in the wrong place.
- Jeremy