[Pixman] [PATCH 0/5] ARMv6: New fast path implementations that utilise prefetch

[Søren Sandmann]

Ben Avison <bavison at riscosopen.org> writes:

> On Sat, 12 Jan 2013 13:15:18 -0000, Søren Sandmann <sandmann at cs.au.dk> wrote:
>> Yeah, that's fair enough. The following patch adds some information to
>> the README file - I'd appreciate feedback on whether this type of
>> documentation would have been helpful to you. If so, I'll also add
>> something on pixman.org
>
> Hi Søren,
>
> Yes, that does look useful - I can spot a few things that I eventually
> deduced by googling, and one or two things that I still hadn't picked up
> on. Here are a few comments:

Thanks for the comments.

> * You might want to use an overview URL for that Bugzilla database, like
>   https://bugs.freedesktop.org/buglist.cgi?product=pixman&component=pixman&resolution=---

Well, this section is intended for people who wants to file bugs, and no
so much for developers. I suppose such an overview could be useful for
people who want to check if their bug is already known, but we get few
enough bug reports that marking them as duplicates isn't a significant
burden.

> * You might want to make it more clear that git send-email must be used to
>   submit patches, in preference to most email clients that will mangle
>   them, or the use of attachments. I recognise that you might consider
>   this outside the scope of your document, but I found a number of things
>   non-obvious about git-email: for example, it's distributed in a
>   different package than git, and despite what git-scm.com says, you don't
>   need access to an IMAP server to be able to use it, but you can use a
>   locally-installed SMTP server such as msmtp instead (though setting that
>   up wasn't that easy either, for reasons I won't go into here).

The new patch doesn't mention format-patch anymore, only send-email.

> * The --cover-letter option to git format-patch you mentioned: I assume
>   this is for the [PATCH 0/n] emails. It's probably worth mentioning that
>   most of the detail of benchmarking etc still needs to go into the
>   individual patches. Presumably this is because the cover letter doesn't
>   get imported by the "git am" command?

I removed the part about --cover-letter in favor of a guideline that
says that big patch series should have an introductury mail. There is no
real need for people to use --cover-letter.

> * Siarhei mentioned that the source tree is expected to continue to build
>   between each part of a multi-part patch series. Implicit in this is the
>   fact that each part can contain patches for more than one source file
>   (which wasn't immediately obvious to me).

This is really part of "working knowledge of git". The reality is that
someone who is unfamiliar with git has to get up to speed with it before
he can contribute more than simple patches.


Soren