This is the mail archive of the
gdb@sourceware.org
mailing list for the GDB project.
Re: Patchwork patch tracking system
- From: Eric Christopher <echristo at gmail dot com>
- To: Joel Brobecker <brobecker at adacore dot com>
- Cc: Gary Benson <gbenson at redhat dot com>, Stan Shebs <stanshebs at earthlink dot net>, "gdb at sourceware dot org" <gdb at sourceware dot org>
- Date: Tue, 22 Apr 2014 08:40:17 -0700
- Subject: Re: Patchwork patch tracking system
- Authentication-results: sourceware.org; auth=none
- References: <20140402100842 dot GA956 at blade dot nx> <533F3713 dot 40700 at earthlink dot net> <20140417135040 dot GA891 at blade dot nx> <20140422130652 dot GG5790 at adacore dot com>
On Tue, Apr 22, 2014 at 6:06 AM, Joel Brobecker <brobecker@adacore.com> wrote:
>> > So if we try it and like it, how does one go about transitioning
>> > from "trial" to "real"?
>>
>> I guess by the people doing the reviewing deciding to use it.
>> It may be it is useful even with only a subset of reviewers
>> using it. I can't determine this myself, I need feedback from
>> people who are reviewing regularly.
>
> In my opinion, the GDB project is in dire need of a way to track
> patches. Using one's mailbox to track patches just does not work.
> But I think that we would need full commitment to the tool from
> the project, or else it'd quickly start overflowing with stale
> info.
>
> There is a tool that we use internally at AdaCore which I was starting
> to think of proposing for GDB, called geritt. From what I have been
> able to see from patchwork's webpage, geritt seems like a much more
> advanced system compared to patchwork. But the tradeoff is that using
> geritt requires a bit more work as well, and that part or all of
> the review process would happen on geritt, rather than the mailing-list.
> It's not very intuitive at first, but it is very easy and lightweight.
>
> I personally believe geritt's approach to be better in the long run.
> But, while I am worried about having communication and patch handling
> be done via two distinct systems, patchwork's simpler approach might be
> working well enough without requiring the big shift in patch-reviewing
> paradigm.
>
FWIW we (some of the google folk) looked at geritt for LLVM and
discarded in favor of phabricator. It seemed to solve a lot of the
problems that we had and allowed communication to and from the mailing
lists for patches which was key for us as we have a similar review
style to gcc/gdb/binutils. We didn't want to remove the ability for
people to send patches to the mailing lists, but yet get a better
review mechanism for large patches/queuing/etc.
Just piping in since we recently did some of this work. Feel free to
let me know if you have any questions on our experiences.
-eric