Re: Incorrect Author on patch

[Laszlo Ersek <lersek@redhat.com>]

On 04/25/18 17:02, Evan Lloyd wrote:
> Hi Ruiyu.
> When we look at the edk2 git log, or GitHub https://github.com/tianocore/edk2/commit/ee4dc24f57c32a445e7c747396c9bfbd8b221568
> we see that Sami's AcpiView patch shows you as the Author.
> I'm not sure what might have caused that (and it is obviously accidental), but it is a little unfortunate in that the commit doesn't show up on Sami's GitHub stats.
> Fortunately, this is not currently significant for our organisation, although I'm sure Sami would prefer to have the credit.
> This is a trivial matter, but you may wish to understand what caused it so that you don't accidentally upset someone for whom the stats are significant.

Right, this occurred at least one other time as well: see commit
8b5c80e0296c ("MdeModulePkg/UefiBootManagerLib: fix
AddLoadOptionVariable docs/prototype", 2018-04-23). Ray pushed the patch
(so the Committer field is certainly right), but Ray's name+email
replaced mine in the Author field. I had noticed that earlier, but now
I'm seeing it as a pattern.

I believe this is a tooling issue. I notice the following bit on the
commit message:

    Change-Id: I8a609d6502b6f8929b2f568acaa147065003b6f4

I certainly didn't post the patch with that, and I doubt Ray added it
manually. So, whatever tool Ray used to handle the patch lost the
authorship information.

And, I see the exact same kind of tag, namely

    Change-Id: Ifa23dc80ab8ab042c56e88424847e796a8122a7c

on commit ee4dc24f57c3 ("ShellPkg: Add acpiview tool to dump ACPI
tables", 2018-04-23), which you mention.

... In fact, my patch happens to be the direct ancestor of Sami's, in
the git history, and their commit times are just ~3 minutes apart. I'm
quite certain Ray has started using a new tool.

For example, commit bc2288f59ba2 ("UefiCpuPkg/MpInitLib: put
mReservedApLoopFunc in executable memory", 2018-03-08) was also
committed by Ray, *but* Jian's authorship was preserved fine. (No
"Change-Id" either.)

... Oh... it looks like those Change-Id's were added by Gerrit:

https://git.eclipse.org/r/Documentation/user-changeid.html

And then, please look at this:

https://gerrit-review.googlesource.com/Documentation/error-invalid-author.html

    For every pushed commit Gerrit verifies that the e-mail address of
    the author matches one of the registered e-mail addresses of the
    pushing user. If this is not the case pushing the commit fails with
    the error message "invalid author". This policy can be bypassed by
    having the access right 'Forge Author'.

Uh, what?... Gerrit says "invalid author" if Ray pushes a patch that
wasn't authored by himself? And the option to override that is called
"forge" author? O_o

Anyway, the page continues,

    If pushing to Gerrit fails with the error message "invalid author"
    and somebody else is author of the commit for which the push fails,
    then you have no permissions to forge the author identity. In this
    case you may contact the project owner to request the access right
    '+1 Forge Author Identity' in the 'Forge Identity' category or ask
    the maintainer to commit this change on the author’s behalf.

Ray, if you absolutely must use Gerrit, please make sure that you have
the '+1 Forge Author Identity' access right. IMO, we maintainers are
responsible for preserving git metadata the best we can.

(For example, if a patch is applied from the mailing list with "git am",
it preserves the authorship information -- the documentation says, 'The
commit author name is taken from the "From: " line of the message, and
commit author date is taken from the "Date: " line of the message.')

Thank you!
Laszlo