- Jan 11, 2018
-
-
Signed-off-by:
Gustavo Padovan <gustavo.padovan@collabora.com> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch>
-
Signed-off-by:
Jani Nikula <jani.nikula@intel.com> Signed-off-by:
-
- Jan 10, 2018
-
-
Simona Vetter authored
Since we still have git gc fighting with our own gc algo it can happen that git removes cached rerere entries before we decide to do. This is mostly the case for unresolved conflicts (since there git deletes them after 15 days already, while we keep them for 60 - the default only for resolved conflicts). To avoid confusion simply ignore this case. pull_rerere_cache should clean up the mess if necessary. v2: Fix typo while at it (Archit) Acked-by:
Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by:
Archit Taneja <architt@codeaurora.org> Reported-by:
Daniel Vetter <daniel.vetter@intel.com>
-
- Dec 19, 2017
-
-
Previously 'dim sparse' just touched all the files affected by the range and run sparse against that. Since this may show issues from included headers and sparse-compliance does not seem to be strongly enforced the output is rather unhelpful. With this change the repo is checkout at each commit, sparse is run on incrementally, on every file that was changed. Each result is compared to it's parent's result using remap-log to adjust for potential line numbering changes. Only the difference is reported. v2: Clearer wording on remap-log check. Use the incremental build. Reset ref on exit. v3: Make shellcheck happy: * Ignore SC2064 in trap * Quote array expansion Signed-off-by:Arkadiusz Hiler <arkadiusz.hiler@intel.com> Acked-by:
-
remap-log parses patches to build a map of changes in the source files. The tool can use that map to remap line numbers in log files, so the logs are easily diffable. Signed-off-by:
-
Simona Vetter authored
Ever since we implemented the explicit garbage-collection logic for the rr-cache/ entries there's been reports by committers that the git pull in drm-rerere failed. I think I've seen that now enough times to distill at least 2 patters: - Race with someone else hitting the same conflict, or at least a conflict with the same conflict-sha1. Then you have locally preimage (and perhaps other files) which conflict. Because those files are numbered it's pretty hard to merge them automatically, so best to just throw away the local stuff. - git gc removed a bunch of rr-cache/ entries. We've never figured out how exactly the gc logic for rr-cache entries works, except it seems to be atime based and extremely erratic. Again best option is to just reset and let dim gc rr-cache/ entries explicitly (and in a much more consistent way). I think the only risk in this patch is if 2 people resolved the exact same conflict in drm-tip, but then maybe they should start talking with each another on irc instead of just looking at dim output. Throwing away the 2nd conflict resolution shouldn't result in much harm at least. Reviewed-by:
-
Simona Vetter authored
Just prep for the next patch. Reviewed-by:
-
- Dec 14, 2017
-
-
Simona Vetter authored
This fixes commit 8ea00c70 Author: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Date: Thu Dec 7 11:42:28 2017 +0200 dim: Move to more readable fixup patch naming scheme Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by:
-
- Dec 13, 2017
-
-
Simona Vetter authored
Acked-by:
-
Simona Vetter authored
Since we rotate maintainer functions now on a per-release basis, and that seems to be a superior model (bettter exchanged and sharing of knowledge) per-maintainer templates don't make much sense. Instead use per-branch templates. To keep some separation between the scripts and the setup I opted to put the templates into the drm-rerere branch, where we have nightly.conf already. I've pushed a few examples there of what I have in mind. Once this has landed I'll send out the patch for drm-rerere with some templates to start with. v2: Align to the new file layout as discussed with Joonas. Comment for function parameters. Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Acked-by:
-
Simona Vetter authored
Just prep work. Well, result is that the tag template is also used for drm-intel-next tags. Acked-by:
-
Simona Vetter authored
dim apply auto-adds this now anytime it's needed. v2: Fixup cat-to-fixup completion Acked-by:
-
Simona Vetter authored
Oops Acked-by:
-
Joonas Lahtinen authoredPut yellow highlight on drm-tip to point out that it is a shared tree. Signed-off-by:
Jani Nikula <jani.nikula@linux.intel.com>
-
Joonas Lahtinen authored
Because branch and repo already contain '-', move to use '__' as the delimiter between the two for more readability. While at it, allow explicitly specifying the branch where the fixup should be applied in the case of a non-conflicting merge issue. Fixing the merge commit of drm-misc-next after noticing drm-tip not to build would be done in the following way: $ git diff | dim cat-to-fixup drm-misc-next v2: Bash completion v3: Use "fixups/" folder under rerere. Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by:
-
- Dec 12, 2017
-
-
Simona Vetter authored
dim push doesn't check everyone's patches (since most subsystems don't have strict review requirements). But when we take a pull request we really want to make sure people follow due process. Motivated by the recent gvt pull, which lacked a commiter sob. Reviewed-by:
-
- Nov 28, 2017
-
-
Simona Vetter authored
Requested by Arek (so CI can use sparse/checkpatch) and Chris (because it's more useful for developers this way). v2: Remove debug leftover. Oops (Chris). v3: dim retip requires the integration configuration to figure out where drm-tip is: - Read the integration config even for developer cmds, but only when available. - Add a fallback trick for dim retip to make a good educated guess about where drm-tip is. v4: Add break (Chris, Gustavo) Acked-by:
-
Simona Vetter authored
I want to add more here. Also fix an issue where you could never see the usage hint if you typed the wrong subcommand without setup being done yet. Acked-by:
-
- Nov 27, 2017
-
-
Simona Vetter authored
To make sure the report is actually accurate. Motivated in a discussion with Joonas. Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
-
- Nov 20, 2017
-
-
Jani Nikula authored
Take into account git commit e66d7c37a5f1 ("request-pull: capitalise "Git" to make it a proper noun"). Reported-by:Zhenyu Wang <zhenyuw@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by:
-
- Oct 12, 2017
-
-
Jani Nikula authored
Use a timestamp file to check for updates every 24 hours. Acked-by:
-
Jani Nikula authored
time was local in dim_rebuild_tip, so not too nice to reference in functions called from it. Reported-by:
Daniel Vetter <daniel@ffwll.ch> Acked-by:
-
Jani Nikula authored
Seems like a batter name for a global. Acked-by:
-
- Oct 11, 2017
-
-
Jani Nikula authored
Make it easier to debug potential issues in rerere cache handling. Adding the commit message using a temp file also makes it easier to add more breadcrumbs in the future. Very-much-acked-by:
-
Jani Nikula authored
Bail out early on errors to reduce indent. Add local decls while at it. Reviewed-by:
-
Simona Vetter authored
Otherwise the tree rebuilding fails. This might be the reason why some git versions create different conflicts, since checking git-sources/rerere.c didn't provide any hint at all about what has changed. Note this is only --local, i.e. for the current git repo, not --global or --system. But with worktrees that includes a lot more than just the drm-tip repo. But then we just force the default, so hopefully the impact is minimal. Cc: Maxime Ripard <maxime.ripard@free-electrons.com> Cc: Noralf Trønnes <noralf@tronnes.org> Cc: Chris Wilson <chris@chris-wilson.co.uk> Tested-by:
Maxime Ripard <maxime.ripard@free-electrons.com> Acked-by:
-
Jani Nikula authored
Broken since introduction in cad37e19 ("dim: auto-add remotes"), the remote asked from the user gets ignored. Get the remote name to the remote variable. Also bail out on non-zero returns from read. Cc: Daniel Vetter <daniel@ffwll.ch> Acked-by:
-
- Oct 09, 2017
-
-
Jani Nikula authored
While at it, drop an old drm-intel reference. Reviewed-by:
-
Jani Nikula authored
No reason for the function to be so specific, make it more flexible in case we need to choose ssh URLs later on. Also rename it appropriately. Remove the fallback of converting ssh:// URLs to git:// URLs now that we have all the URLs with all the accepted protocols defined in nightly.conf since drm-rerere commit 7aa4d6332234 ("nightly.conf: add url lists to all repos"). Reviewed-by: -
Jani Nikula authored
We have a helper for this, use it. Reviewed-by:
-
Jani Nikula authored
dim setup is now the only place that may use them. Going forward, dim setup could go one step further and dig out nightly.conf directly from the repo before we've created a worktree checkout for it. Reviewed-by:
-
Jani Nikula authored
Since drm_tip_repos now contains drm-tip, we can use it to reduce dependency on URLs hard-coded in dim. Reviewed-by:
-
Jani Nikula authored
Since commit 1c228d7de23c ("nightly.conf: add drm-tip and linux-upstream to drm_tip_repos") in drm-rerere, nightly.conf has drm-tip and linux-upstream. Don't handle them specially. Reviewed-by:
-
- Oct 06, 2017
-
-
Jani Nikula authored
Don't assume current directory works. Reviewed-by:
-
Jani Nikula authored
Look up the remotes using the repo names in nightly.conf. Reviewed-by:
-
Jani Nikula authored
We have all the plumbing necessary to figure out the drm-intel remote. No need to configure it separately. Reviewed-by:
-
Jani Nikula authored
The maintainer repo hasn't been Intel specific for a while now. Rename the variable. DIM_DRM_INTEL is used as a default for DIM_REPO for backward compatibility. Reviewed-by:
-
Jani Nikula authored
The hard-coded list of remotes and branches has gone stale. Reviewed-by:
-
Jani Nikula authored
Since commit cad37e19 ("dim: auto-add remotes") url_to_remote can't really fail. Same for repo_to_remote when the repo exists. Redirecting their output when the remote isn't there leads to url_to_remote waiting for user input without prompting, giving an appearance of a hang. While at it, change the exit to a return. set -e at the top takes care of aborting. Using git worktrees is now practically required because we need all the remotes everywhere. Worktrees have been the default for a while now, and you have to go out of your way to set dim up without worktrees. You can still get by without, but it'll be clumsy. Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Sean Paul <seanpaul@chromium.org> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by:
-
Jani Nikula authored
No longer consider drm-intel as the default. All repos need to be explicitly specified. Improve error reporting in repo_to_remote for unknown repos. Reviewed-by:
-
