On 9/11/19 5:40 PM, Martin K. Petersen wrote:
> * Do not use custom To: and Cc: for individual patches. We want to see the
> whole series, even patches that potentially need to go through a different
> subsystem tree.
Thanks for having written this summary. This is very helpful. For the
above paragraph, should it be clarified whether that requirement
applies to mailing list e-mail addresses only or also to individual
e-mail addresses? When using git send-email it is easy to end up with
different cc-lists per patch.
I prefer to have the entire series sent to linux-scsi or
target-devel. It wouldn't be so bad if discussions about the merits of a
tree-wide change consistently happened in responses to the cover
letter. But more often than not discussion happens in response to a
patch touching a different subsystem and therefore in a mail exchange
that doesn't end up on linux-scsi.
> * The patch must compile without warnings (make C=1
> and does not incur any zeroday test robot complaints.
How about adding W=1 to that make command?
How about existing drivers that trigger tons of endianness warnings,
e.g. qla2xxx? How about requiring that no new warnings are introduced?
This was in response to a driver submission (for a different driver)
around the time this doc was written. The problem is that it's sometimes
hard to distinguish new warnings from old ones. I'm all for requiring
that no new warnings are introduced.
> * The patch must have a commit message that describes,
> comprehensively and in plain English, what the patch does.
How about making this requirement more detailed and requiring that not
only what has been changed is document but also why that change has
I'd really like all this patch submission guideline material to live in
Documentation/process. But yes.
Martin K. Petersen Oracle Linux Engineering