Skip to content
Snippets Groups Projects
Select Git revision
0 results
Search by author
  • Any Author
0 authors
  1. Mar 04, 2025
    • Rodrigo Vivi's avatar
      dim: Restore fix_of_fix · cbb4e4a0
      Rodrigo Vivi authored and Rodrigo Vivi's avatar Rodrigo Vivi committed 
      
The original fix of fix on commit 28155089 ("dim: Consider
fix-of-fix for -fixes cherry picking.") was:

fix_of_fix=$(git log --grep="cherry picked from commit $fixes" \
		  --after=6months --format=format:%h \
		  $intel_remote/drm-intel-next-fixes -1)
if [[ -z "$fix_of_fix" ]]; then
 fix_of_fix=$(git log --grep="cherry picked from commit $fixes" \
 --after=6months --format=format:%h $intel_remote/drm-intel-fixes -1)

But commit e7f73056 ("dim: Generalize cherry-pick fixes")
generalized it to for b in drm-intel-next drm-intel-gt-next

So, we need to get back to look for the commits in the fix branch
not in the next branch. However, let's only look to the current
fix branch instead of both fixes branches, otherwise we will can
reach a point where we can try to cherry-pick a fix_of_fix to a branch
which doesn't have the first-fix patch.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: default avatarRodrigo Vivi <rodrigo.vivi@intel.com>
      cbb4e4a0
  3. Feb 07, 2025
  5. Sep 19, 2024
  7. Sep 03, 2024
  9. Jul 29, 2024
    • Lucas De Marchi's avatar
      dim: display changes are under drm-intel maintenance · 9e645341
      Lucas De Marchi authored 
      
Let's keep merging display stuff through drm-intel when possible as
there's less chance of conflicting.

Signed-off-by: default avatarLucas De Marchi <lucas.demarchi@intel.com>
      9e645341
    • Lucas De Marchi's avatar
      dim: Drop fetch in dim_cherry_pick · 7b537a01
      Lucas De Marchi authored and Lucas De Marchi's avatar Lucas De Marchi committed 
      
The fetch is currently called with an unbound `remote` variable,
which results in fetching the upstream remote of the current branch.
In most cases this is the correct behavior as the branch we are
cherry-picking changes in should have the upstream set. However if
the upstream is not set, it will a fetch origin, which has nothing
to do with the operation.

Just drop the fetch as it's not needed for 2 reasons:

1) If the user is giving commit X, he probably have that commit locally
   already and no fetch is needed
2) When listing the references, we will fetch drm-tip, which should have
   all commits we are possibly cherry-picking

Signed-off-by: default avatarLucas De Marchi <lucas.demarchi@intel.com>
      7b537a01
  11. Apr 19, 2024
  13. Apr 18, 2024
    • Lucas De Marchi's avatar
      dim: tag once, try 2 · 0daf0be5
      Lucas De Marchi authored 
      
Tag just once to avoid the old tag left on the local machine with
possibly the wrong message. Even more importantly: the revert of first
version was correct, but at least with my current git version (2.44) it
doesn't work. The second call to git-tag to amend the commit message
will also move the tag to another commit:

	+ git tag -F- drm-xe-fixes-2024-04-18-1 'drm-xe-fixes@{upstream}'
	+ '[' git == git ']'
	+ key_arg=-s
	+ git tag -s -f drm-xe-fixes-2024-04-18-1
	Updated tag 'drm-xe-fixes-2024-04-18-1' (was 46d02aa1b02c)

So, even if it showed the correct log in gitk and updated the message,
it ended up tagging the wrong commit.  We could add "$branch@{upstream}"
in the second invocation, but I think this also highlight another
shortcoming of calling git-tag twice.

In this "try 2" to avoid the breakage from v1 with some editors, do not
pass the summary as stdin to `git-tag`, but rather create a tmp file.

Signed-off-by: default avatarLucas De Marchi <lucas.demarchi@intel.com>
      0daf0be5
  15. Apr 17, 2024
    • Rodrigo Vivi's avatar
      Revert "dim: tag once" · f5fb02e0
      Rodrigo Vivi authored 
      The -e usage breaks EDITOR="emacs -nw":

hint: Waiting for your editor to close the file... emacs: standard input is not a tty
error: There was a problem with the editor 'emacs -nw'.
Please supply the message using either -m or -F option.

And simply removing the -e tag didn't worked out. It then
simply goes there and create the tag with the template without
stopping on emacs.

This reverts commit b8409429.
      f5fb02e0
  17. Apr 15, 2024
  19. Apr 11, 2024
  21. Mar 27, 2024
  23. Mar 26, 2024
  25. Mar 05, 2024
  27. Mar 04, 2024
  29. Feb 29, 2024
    • Jani Nikula's avatar
      dim: add support for applying patches using b4 · 4dc7d0cc
      Jani Nikula authored 
      Add 'b4-shazam-branch' and 'b4-shazam' subcommands to apply a patch
series using b4 [1], fetching the patches from Lore. You can either
specify the series using message-id on the command-line, or by piping a
message in the series to the subcommand. Additional b4 command-line
options may be specified as needed, for example to apply individual
patches.

b4 is a much more robust alternative to pretty much everything we do in
dim_apply_branch() and apply_patch() functions, and automatically
gathers reviews and acks from the list.

Use the b4.trailers-ignore-from config to ignore trailers from CI and
kernel build bot. b4 v0.13 or later is required to filter all trailers
in messages from the mentioned addresses (for example Fixes:
trailers). Otherwise b4 shazam picks up a lot of garbage trailers from
CI mails.

[1] https://b4.docs.kernel.org/en/latest/index.html



v3:
- Unconditionally require b3 v0.13 or later
- Use --config option instead of env vars
- Add DIM_B4 config option

v2:
- Use GIT_CONFIG_* to set b4.linkmask (Lucas)
- Use GIT_CONFIG_* to set b4.trailers-ignore-from
- Check for b4 command
- Check b4 version for cover trailers
- Use subcommand names b4-shazam and b4-shazam-branch

Signed-off-by: default avatarJani Nikula <jani.nikula@intel.com>
      4dc7d0cc
  31. Feb 28, 2024
  33. Feb 27, 2024
    • Maxime Ripard's avatar
      dim: Handle drm.git move to Gitlab automatically · 9ab59fc4
      Maxime Ripard authored 
      
As part of the transition of DRM from cgit to Gitlab, we updated
drm-rerere's nightly.conf file to reflect the new URL.

However, dim updates drm first and drm-tip lasts, so if you update the
git remote URL like we instructed before running dim update-branches (or
any command that fetch all remotes), it will error out before it has the
chance to retrieve the nightly.conf file.

If the user doesn't update the git remote URL, then dim update-branch
will run, fetch drm through the cgit repo, update nightly.conf. On the
second run, it will detect that the URL doesn't match anymore, but since
there's already a drm remote setup, it will fail to add the remote.

So, in either case, dim's broken.

Add a bit of logic that detects the drm remote URL and will update
remote url if we're using our legacy URLs.

Co-developed-by: default avatarDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: default avatarDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Maxime Ripard's avatarMaxime Ripard <mripard@kernel.org>
      9ab59fc4
    • Lucas De Marchi's avatar
      dim: make dry-run a bit more useful for pull requests · 88cd73ca
      Lucas De Marchi authored 
      dim -d pull-request was failing with:

	Using drm/drm-next as the upstream
	../maintainer-tools/dim: line 2225: gitk: command not found
	git tag -F- drm-xe-next-2024-02-27 drm-xe-next@{upstream}
	git tag -a -f drm-xe-next-2024-02-27
	error: src refspec drm-xe-next-2024-02-27 does not match any
	error: failed to push some refs to 'ssh://gitlab.freedesktop.org/drm/xe/kernel.git'

Make it a little bit more useful. Probably we shouldn't skip that much,
but that can be left for later. Same command now shows:

	Using drm/drm-next as the upstream
	../maintainer-tools/dim: line 2229: gitk: command not found
	git tag -F- drm-xe-next-2024-02-27 drm-xe-next@{upstream}
	git tag -a -f drm-xe-next-2024-02-27
	To ssh://gitlab.freedesktop.org/drm/xe/kernel.git


	 * [new tag]                     b2121f2bd2232cd0556b2182078d159d81497885 -> drm-xe-next-2024-02-27
	Skip.
	neomutt -s [PULL] drm-xe-next -i /tmp/tmp.eu5PC10boN -c Jani Nikula <jani.nikula@linux.intel.com> -c Joonas Lahtinen <joonas.lahtinen@linux.intel.com> -c Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> -c Rodrigo Vivi <rodrigo.vivi@intel.com> -c Thomas Zimmermann <tzimmermann@suse.de> -c Maarten Lankhorst <maarten.lankhorst@linux.intel.com> -c Maxime Ripard <mripard@kernel.org> -c Thomas Hellström <thomas.hellstrom@linux.intel.com> -c Oded Gabbay <ogabbay@kernel.org> -c Lucas De Marchi <lucas.demarchi@intel.com> -c dri-devel@lists.freedesktop.org -c intel-gfx@lists.freedesktop.org -c intel-xe@lists.freedesktop.org -c dim-tools@lists.freedesktop.org Dave Airlie <airlied@gmail.com> Daniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: default avatarLucas De Marchi <lucas.demarchi@intel.com>
      88cd73ca
    • Lucas De Marchi's avatar
      dim: Pull requests to use https:// · b7cc9a0c
      Lucas De Marchi authored 
      Not all repos have a git:// protocol enabled. Once upon a time the
justification to use it was due to bad performance of the https backend.
https://git-scm.com/book/en/v2/Git-on-the-Server-The-Protocols

 still
marks it as "the best on performance", but with the Smart HTTP backend,
the cons start to have a bigger role. Also worth noting that it's not
available on Github and Gitlab anymore due to security issues.

Signed-off-by: default avatarLucas De Marchi <lucas.demarchi@intel.com>
      b7cc9a0c
  35. Feb 26, 2024
  37. Feb 01, 2024
  39. Jan 29, 2024
  41. Jan 23, 2024
  43. Jan 11, 2024
  45. Jan 10, 2024