Overview
Request 1191638 accepted
please don't accept
- Created by rrahl0
- In state accepted
-
Package maintainers:
ojkastl_buildservice and
rrahl0
Loading...
openSUSE Build Service is sponsored by
please don't accept
ojkastl_buildservice and rrahl0
So I know this is quite radical, but I wanted to hear your opinion on this.
About the _service file: I do like to rename the file to %{pkg_name} and .git gets autoexcluded. using refs/tags/%{version} helps imo to specify that it is actually a tag, not a branch.
about the spec file itself, I added podman/docker with service file and fixed the wrong zsh completions directory. additionally I removed the Requires: completions as it's not necessary to include this. and if somebody wants to install the completions without the package itself, so let him do that :3.
Removed the Group: line as rpm deprecates it in the future anyway (which should happen in rpm 4.20).
Using
refs/tags/...sounds like a good idea. Using the filename is neat, I tried this years ago and failed so I just kept using the basename or archive for other services.I tend to keep the exclude for
.gitto make it clear for unexperienced readers that this is being excluded.hm, wouldn't it make it more clear to use package-meta=no?
IMHO on the contrary, who understands what this options means? If it were "package-metadata" or "package-scm-metadata" at least, but package-meta is a misnomer.
I mean I would add a comment about that this is the git metadata. on the other hand when excluding .git, people would think that removing that line actually adds it
It spits out warnings on SLE/Leap 15.x and did not harm yet. So I had it in my copy&paste templates for completion subpackages. But I won't object to removing it.
Thanks, I'll take a closer look and see what you did.
Huh? I have that in dozens of other packages, so either both are working or there not many zsh users actually using the completions... :-)
I think it is just good packaging style to add them. Of course, allowing the user to shoot herself in the foot is nice, but those lines do not harm.
idk, I only ever met one person who installed the -bash-completions to actually install the package. and zypper autoinstalls the bash completions when you install the packge. (as long as bash-completion is installed)
on the topic of zsh. when rpm -ql the package that directory doesn't exist. and my own package (tailscale) has is somewhere else
I'll write a mail to the packaging list to clarify this. If there really is a wrong directory, I have that in dozens of my packages...
Uuuh, forgot the most important things:
Thanks for checking and taking the time to come up with this! :-)
I started a discussion last year (on the packaging or go list) but this lead nowhere and I got no result, but apparently this line was necessary for $REASONS long ago. No one know if it is still needed...
I have never seen this, and even the go_nostrip is not used anymore
Those are not necessary.
these are required because I was told my multiple people (can't remember who) that a directory should never be owned by multiple packages. which means you need them, so the directories are owned (and the package builds)
this is why fedora has -filesystem packages
I was told some years ago that zypper does no longer throw a fit if two packages own the same directory. :-)
But I won't object to adding them, does not make a big difference in the builds duration I guess.
tbf, if I have the possibility to choose, I would rather have not 2 packages owning one directory ;). less stuff can break
As I said, I would not object.
I'll go through your changes this evening and cherry-pick/accept them according to our discussions above.
Wanna co-maintain this package with me?
sure, I do actually like maintaining the whole stack of most stuff I maintain, as otherwise you get always in some issues with "you need to update package x before your package can get updated"