Re: [edk2-devel] [PATCH] MdePkg/DxeHstiLib: Added checks to improve error handling.

["Laszlo Ersek" <lersek@redhat.com>]

On 08/29/19 19:02, Ni, Ray wrote:
> 
> 
>> -----Original Message-----
>> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Laszlo Ersek
>> Sent: Thursday, August 29, 2019 7:24 AM
>> To: Leif Lindholm <leif.lindholm@linaro.org>; Ni, Ray <ray.ni@intel.com>
>> Cc: Gao, Liming <liming.gao@intel.com>; Cetola, Stephano <stephano.cetola@intel.com>; Kinney, Michael D
>> <michael.d.kinney@intel.com>; Jayanth.Raghuram@dell.com; 'Andrew Fish (afish@apple.com)' <afish@apple.com>;
>> Wei.G.Liu@dell.com; devel@edk2.groups.io
>> Subject: Re: [edk2-devel] [PATCH] MdePkg/DxeHstiLib: Added checks to improve error handling.

>> - Doesn't work for patch series, only for single patches.
> Laszlo,
> Do you mean attachment is not allowed for a series of patch?

That's correct; I mean that.

> Why? A mail can carry multiple attachments. That makes the patch series easy to fetch in my opinion from Outlook.

The goal is to enable fine-grained, email based review comments.
Meaning, a reviewer responds to the original email, whereby the original
patch is quoted in full in the response. Then the reviewer inserts
comments near the code locations in the patch that need updates.

This approach works perfectly when the original email was sent with
git-send-email.

The approach works tolerably (as an exception to our workflow) when the
original email was sent manually, with a single patch pasted or
attached. Because, in this case, the review comments can still be made
context-sensitively, and the threading structure will be sane.

The approach does not work at all when a single email carries multiple
patches. Quoting becomes a mess, review comments are a lot more
difficult to associate with the targeted patch within the series, and so on.

If the contribution is a patch series (i.e., not a single patch), and
the submitter is unable to use "git-send-email", then the submitter is
responsible for manually establishing the exact same shallow thread
structure that git-send-email would. Specifically, the submitter first
needs to send a cover letter email, then they need to send each patch
individually, with numbered subjects, in direct response to the cover
letter email. They can paste or attach one patch per email, until the
full series is posted.

The email thread must reflect the git commits so that a specific git
commit has a matching "entry point" into the mailing list discussion --
i.e., a reviewer interested in a given git commit can find the precisely
matching email, and the related *sub*-thread that discusses that
particular commit.

Email is not merely a dumb carrier for patches where it's OK to drop a
single patch-bomb email (let alone a ZIP file) on reviewers. A
well-structured email thread is vital to careful review, and for the
long-term archival of the discussion.

Thanks,
Laszlo