[PATCH] dim: allow to add link to the current HEAD

[Jani Nikula]

On Mon, 18 Nov 2019, Lucas De Marchi <lucas.demarchi at intel.com> wrote:
> On Mon, Nov 18, 2019 at 12:16:19PM -0800, Rodrigo Vivi wrote:
>>     On Nov 18, 2019, at 12:14 PM, Lucas De Marchi
>>     <[1]lucas.demarchi at intel.com> wrote:
>>     On Mon, Nov 18, 2019 at 04:57:34PM +0200, Jani Nikula wrote:
>>
>>       On Fri, 15 Nov 2019, Lucas De Marchi <[2]lucas.demarchi at intel.com>
>>       wrote:
>>
>>         Add method add-link-head that applies the link to head without any
>>         additional check. This is useful either to reduce typing or to add a
>>         link while in the middle of an interactive rebase.
>>
>>       I see the patch does what it says on the box. However:
>>
>>       1) Most commands expect you to supply the branch name so you know what
>>        you're doing. There's tab completion to aid you, but you need to see
>>        what branch you're operating on. This has been in place since there
>>        was only one gun and two feet; now there are quite a few more people
>>        with opportunities to shoot themselves in the foot.
>>
>>       2) I think we should discourage pushing locally rebased
>>        patches. Recommend only pushing patches that apply cleanly. I'd go as
>>        far as removing the add-link* subcommands altogether.
>>
>>     Not sure how this is helping or avoiding that. "push" was never
>>     mentioned here and it's not what this is about.
>>
>>     This was useful for me while rebasing a "feature branch"
>>     to prep a new revision.  It's useful for the future developer reading
>>     the patch to have a link to the revision where the patch was discussed,
>>     rather than a link to a last revision where all the reviews were already
>>     collected.
>>
>>   Very good point. But we need to do in a way that we don't loose the
>>   last link. The last one is the proof that it passed CI.
>
> If you later apply the patch using, say, "dim aq" a second link will be
> added, the link that is already there won't be replaced.
>
> I think the link to the review is much more useful than the link where
> CI reported a pass. Either because of CI flip flops or because a patch
> that passed CI being applied on top of 10 additional commits doesn't
> prove anything.

I think the Link: tag should point at the patch on the mailing list that
was applied. You should only apply patches that passed CI. This is
unambiguous and does not leave any subjective wiggle room. It provides
the audit trail (for want of a better expression) we want. And more
often than review, I'm actually after the CI results of a commit.

> Anyway, if people just put the automatically-added "Link" when you
> download a patch from patchwork, I will just stop having more work on my
> end and follow that.

I don't download patches from patchwork, I apply them from my mailbox. I
apply the patch that passed CI.


BR,
Jani.


>
> Lucas De Marchi
>
>>
>>     Also any patch being applied to any branch are in other words being
>>     rebased since your branch probably changed from the time you wrote it
>>     and the time it was applied, even if it applies cleanly (your rebase
>>     might have being clean, too).
>>
>>     Lucas De Marchi
>>
>>       BR,
>>       Jani.
>>
>>         Signed-off-by: Lucas De Marchi <[3]lucas.demarchi at intel.com>
>>         ---
>>         dim | 23 +++++++++++++++--------
>>         1 file changed, 15 insertions(+), 8 deletions(-)
>>
>>         diff --git a/dim b/dim
>>         index 1c2da80..3aad25f 100755
>>         --- a/dim
>>         +++ b/dim
>>         @@ -1255,17 +1255,11 @@ function dim_rebase
>>         fi
>>         }
>>
>>         -function dim_add_link
>>         +function dim_add_link_head
>>         {
>>         -  local branch file message_id
>>         +  local file message_id
>>
>>         -  branch=${1:?$usage}
>>         -  shift
>>         file=$(mktemp)
>>         -
>>         -  assert_branch $branch
>>         -  assert_repo_clean
>>         -
>>         cat > $file
>>
>>         message_id=$(message_get_id $file)
>>         @@ -1279,6 +1273,19 @@ function dim_add_link
>>         fi
>>         }
>>
>>         +function dim_add_link
>>         +{
>>         +  local branch file message_id
>>         +
>>         +  branch=${1:?$usage}
>>         +  shift
>>         +
>>         +  assert_branch $branch
>>         +  assert_repo_clean
>>         +
>>         +  dim_add_link_head
>>         +}
>>         +
>>         function dim_add_link_queued
>>         {
>>         dim_add_link drm-intel-next-queued "$@"
>>
>>       -- 
>>       Jani Nikula, Intel Open Source Graphics Center
>>
>>     _______________________________________________
>>     dim-tools mailing list
>>     [4]dim-tools at lists.freedesktop.org
>>     [5]https://lists.freedesktop.org/mailman/listinfo/dim-tools
>>
>>References
>>
>>   Visible links
>>   1. mailto:lucas.demarchi at intel.com
>>   2. mailto:lucas.demarchi at intel.com
>>   3. mailto:lucas.demarchi at intel.com
>>   4. mailto:dim-tools at lists.freedesktop.org
>>   5. https://lists.freedesktop.org/mailman/listinfo/dim-tools

-- 
Jani Nikula, Intel Open Source Graphics Center