Requested by QA so that when they do stack releases from -testing they
can resurrect them. Also sprinkle a few more $DRY all over the place.

Damien Lespiau authored 10 years ago and Simona Vetter committed 10 years ago

'remove' is not recognized is a slightly older git (1.7.9.5) on a
slightly older distro. Use 'rm' instead, which also work on the git
version listing 'remove' in git remote help (1.8.3.1).

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Except when -f is set to override this. I've wasted enough time
chasing hard-hangs due to over-eager usage of BUG_ONs.

Damien Lespiau authored 10 years ago and Jani Nikula committed 10 years ago

The "usage" text should explain it all. I found, in my quilt series
handling endeavours, that I wanted to be able to shift the prefix
numbers of a patch series.

v2: Use heredoc for usage string, fix second example, use mv -i (Jani)
v3: Don't use a fancy read for usage() (Jani)
    Collect the files to rename in a first pass, or the renaming process
    can interfere with listing the next files to rename
v4: Require bash (Jani)

Looks-good-to: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Because.

v2: Fix up commit message formatting, noticed by Chris.

v3: Use commit -s to avoid the sob-line issues (Jani).

Apparently the automagic replacement of internal references with tags
is lost :(

Also add a todo for adding properly mailing list references
eventually, that seems to be the latest rave among kernel maintainers.

Oops.

Requested by QA so that they have something easy to add to bugs with
all the information.

Failure to git add. oops.

This is useful for the very rare cases where git rerere fails and
can't remember a conflict resolution. Or the rare cases with silent
conflicts.

Since this happens so seldom I always forget the precise patch file
naming convention and waste half an hour debugging dim. Help the
brainless a bit out here.

Also s/git apply/ patch/ since git apply is too picky about unresolved
merges (which occasionally also need to be fixed up).

Use rst2man and man to display if available, otherwise just use cat and
$PAGER.

Use rst2man and man to display if available, otherwise just use cat and
$PAGER.

We want to avoid the merge-window/early -rc craziness and so updated
drm-upstream for -nightly only manually. But that in turn caused
coordination headaches.

Add an optional sha1 to the tree specifier in nightly.conf to fix a
tree to a given revision. That way we can freeze drm-* to the last
version before the merge window and update it at our leisure.

With this fixed we can lift the ban and go back to auto-updating
drm-upstream.

v2: Pimp commit message and drop now outdated comment (Jani)

Much less noise when a bunch of the branches are just fast-forwards.

Useful to slurp in fixup patches.

v2: Do a bit a better check for dirty rerere-branch.

v3: Use -f/--force switch instead. No real getopts parsing though, so
better place it at the right spot!

v4: Unlazy and use the existing getopt stuff.

Damien Lespiau authored 11 years ago and Simona Vetter committed 11 years ago

On its own, "git push origin" will do something different depending on
the the push.default option. So let's speficy the full local:remote
branch here.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

If I only use the local branch and not the remote one
it might be severly out of date.

commit ba7cfc44
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Feb 26 13:52:50 2014 +0200

    dim: add command line option parsing

didn't shuffle all the subcmds correctly.

While at it fix the parameter checking in pull-request-topic a bit -
it checked for $3, then used $2 and checked for $2 when using $3 ...

Damien stumbled over these. Imo not worth to make the setup code more
robust ...

I've screwed this up too often and applied patch fixes to the wrong
patch ...

... and track it in the rr-cache branch. This way the nightly rebuild
code should always have the latest configuration.

v2: Don't automatically update nightly.conf as requested by Jani.
Instead abort if it's not up-to-date.

v3: Unconditionalyl push the rerere-cache branch to make sure any
nightly.conf commits actually go out.

This way the vim shell folding doesn't get confused.

shifting without positional arguments bails out of the script due to
'set -e'.

We already push out any updates, now it goes both ways. Simplifies
collaboration.

Next up is making sure that we use the same sha1s for drm-upstream
branches.

Damien Lespiau authored 11 years ago and Simona Vetter committed 11 years ago

A ':' was missing (should have been ':+'), but I think we want ':-', not
':+' ie we want to substitute if the parameter is undefined or empty:

  http://www.gnu.org/software/bash/manual/bash.html#Shell-Parameter-Expansion



Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Damien Lespiau authored 11 years ago and Simona Vetter committed 11 years ago

This needs to be useremail, not username here.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Damien Lespiau authored 11 years ago and Simona Vetter committed 11 years ago

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>