MozReview workflow changes

[Gregory Szorc]

There were 3 workflow changes to MozReview today that are worth
broadcasting.

1. Support for specifying reviews in commit messages landed. Add "r?gps" or
similar to commit messages to have a reviewer assigned automatically.

2. We now create 1 Bugzilla attachment/flag per commit. Previously, there
was 1 Bugzilla attachment/flag per series. There is no longer a need to
perform a "Ship It" on the parent review request and the "Ship It" button
has been hidden from parent/root review requests because it no longer does
anything meaningful.

3. Submitting a review on a review request without clicking "Ship It" will
clear the r? flag. Previously, the r? flag state lingered and you had to go
to Bugzilla to clear it.

A one-time migration script will run in a few hours to mass convert
attachments/flags in Bugzilla to the new attachment/flag per commit model.
Expect some bug spam.

Please report any issues via bugs and/or in #mozreview.

[Mark Côté]

On 2015-05-27 8:23 PM, Gregory Szorc wrote:
> A one-time migration script will run in a few hours to mass convert
> attachments/flags in Bugzilla to the new attachment/flag per commit model.
> Expect some bug spam.
>

There was a snafu with this script, and we're going to have to delay
running it until next week. We will keep the change in MozReview,
however, since it is beneficial for all new review requests.

This will result in slightly inconsistent states in BMO MozReview
attachments in the interim. If you have an in-progress review request
and push a change, it will create attachments for only the modified
commits and will leave the attachment for the parent review request in
the bug, with any associated review flags. Feel free to obsolete the
parent requests if they are annoying you. The conversion script will
obsolete all remaining parent attachments and create any missing commit
attachments, with appropriate review flags.

You may have received some bugmail about changes to MozReview
attachments; no actual changes were committed, so please ignore them.

Apologies for the inconvenience.

Mark

[Mark Côté]

On 2015-05-27 11:40 PM, Mark Côté wrote:
> On 2015-05-27 8:23 PM, Gregory Szorc wrote:
>> A one-time migration script will run in a few hours to mass convert
>> attachments/flags in Bugzilla to the new attachment/flag per commit model.
>> Expect some bug spam.
>>
> There was a snafu with this script, and we're going to have to delay
> running it until next week. We will keep the change in MozReview,
> however, since it is beneficial for all new review requests.
>
> This will result in slightly inconsistent states in BMO MozReview
> attachments in the interim. If you have an in-progress review request
> and push a change, it will create attachments for only the modified
> commits and will leave the attachment for the parent review request in
> the bug, with any associated review flags. Feel free to obsolete the
> parent requests if they are annoying you. The conversion script will
> obsolete all remaining parent attachments and create any missing commit
> attachments, with appropriate review flags.
>

Sorry, a correction: please feel free to remove r? flags from parent
attachments, but don't obsolete them--they're used by the conversion
script to detect an unmigrated bug.

Mark

[Mark Côté]

Okay one last follow-up: I'm going to modify the script to verify the
full state of all attachments on a bug and ensure that parents are
obsoleted and all commits have attachments. A bit more work, but safer
in the end.

Mark

[Mark Côté]

After a few false starts, the conversion script ran successfully today.
There was only one error, for a bug and associated review-request
series that are in strange states, which we decided to leave alone. If
you notice anything strange in a MozReview-associated bug, please let us
know.

Mark