Re: fedpkg output dir
On Sat, Jan 30, 2021 at 11:37:27AM +0100, Fabio Valentini wrote:
On Sat, Jan 30, 2021 at 9:19 AM Otto Urpelainen <oturpe(a)iki.fi>
wrote:
>
> Jonathan Wakely kirjoitti 29.1.2021 klo 18.22:
> > On 29/01/21 17:04 +0100, Miro Hrončok wrote:
> >> On 29. 01. 21 16:03, Jonathan Wakely wrote:
> >>>
> >>> So if fedpkg clone just added things to .git/info/exclude there would
> >>> be no need to modify every .gitignore file in every repo on every
> >>> active branch.
> >>
> >> That is already the case \o/
> >>
> >>
https://docs.pagure.org/fedpkg/releases/1.37.html#ignore-files-in-a-clone...
> >>
> >
> > Nice! But making 'fedpkg local' unpack into ./build and then build in
> > there still seems sensible, so the excluded patterns would change for
> > that (I don't care about that as I don't use 'fedpkg local',
but it
> > seems like a good suggestion).
> >
>
> Since I got bitten by this, I could try to improve it. Suggestion here
> seems workable to me. So 1) using ./build in 'fedpkg local' and 2)
> adding that directory 'fedpkg clone' excludes. Clearly defined task with
> limited scope.
>
> One question that remains is this: 'fedpkg clone' already does what it
> does and from this discussion we know that many people are using it. If
> the file locations change, changing fedpkg will lead to confusion,
> annoyance and perhaps worse. And while I have not seen any scripts using
> 'fedpkg local', there may be such. Those would break. So perhaps it
> should actually be a new command, maybe 'fedpkg localbuild' (to match
> 'fedpkg mockbuild'), together with documentation update and runtime
> deprecation notice when using 'fedpkg local'.
I don't think this is so important. The name of the build directory changes
with each package version and build architecture, so it would be awkward to
use in a script. I'd just an an option to 'local', something like
--location=cwd|subdir, and maybe initially keep the default unchanged, and
later flip the default once the new paths have become established.
> How does that sound? Particularly to all of you who actually use
'fedpkg
> local'. I got the understanding that while generally users are happy
> with the current behavior, there is no reason why the file generation
> paths could not or should not be made more git friendly.
If you're doing something like this, why not have it match what
"fedpkg mockbuild" already does?
Everything (including rebuilt srpm, built rpms, build logs) goes into
./results_%{name}/%{version}/%{release}/
And that directory is already covered by the default exclude rules
generated by "fedpkg clone".
It would be great to make 'fedpkg local' use the same output dirs are 'fedpkg
mockbuild'.
The difference between the workflows is annoying.
Zbyszek