Re: [PATCH] ci: configure Coderabbit

["Burakov, Anatoly" <anatoly.burakov@intel.com>]

On 10/2/2025 1:54 PM, David Marchand wrote:
> On Thu, 2 Oct 2025 at 12:34, Burakov, Anatoly <anatoly.burakov@intel.com> wrote:
>>
>> On 10/1/2025 4:23 PM, David Marchand wrote:
>>> On Wed, 1 Oct 2025 at 16:12, David Marchand <david.marchand@redhat.com> wrote:
>>>> On Wed, 1 Oct 2025 at 16:04, Burakov, Anatoly <anatoly.burakov@intel.com> wrote:
>>>>> Can you provide a little more details on what exactly you are proposing?
>>>>>
>>>>> On the back of this patch, I installed CodeRabbit in my IDE (VSCode),
>>>>> and did a review of one of my recent patchsets - it's quite interesting,
>>>>> actually, and the comments provided were basic but meaningful, alas the
>>>>> free version is limited to like 1 review per 30 minutes or something so
>>>>> it's a bit limiting. On top of that, I found the tool a lot more usable
>>>>> than GitHub Copilot reviews, which are attached to GH pull requests
>>>>> rather than Git branches, and they take a lot less time to boot, so I
>>>>> feel like this tool has potential.
>>>>>
>>>>> That said, I'm not sure what this patch is supposed to do - is it to
>>>>> have it set up to review patches automatically?
>>>>
>>>> The ovsrobot creates pull requests in its dpdk github repository, and
>>>> Coderabbit and sourcery are invoked on them.
>>>> Look for a link in patchwork, under the name "ci/github-robot-post".
>>>>
>>>> For example, this exact patch got a branch and pr in ovsrobot/dpdk:
>>>> https://github.com/ovsrobot/dpdk/tree/series_36267
>>>> https://github.com/ovsrobot/dpdk/pull/124
>>>
>>> Btw, you'll notice a bit of noise, due to how the PR is created
>>> against ovsrobot/main, and not the exact DPDK main branch at the
>>> moment the patch is submitted on the ml.
>>>
>>>
>>
>> Wow, I didn't know this was a thing! Sourcery actually seems like the
>> more impressive one in terms of summarizing the changes and aiding in
>> review, but those two in tandem look pretty cool indeed. Makes me think
> 
> Well, Robin gave better feedback on Coderabbit than on Sourcery so far.
> Coderabbit seems less picky on credits when it comes to opensource
> projects (I did not check in details, that's *my* feeling when looking
> at reviews in DPDK PRs where I see sourcery complaining we consumed
> too many credits recently).
> 
> One important difference between the two is that it does not seem
> possible to tweak sourcery with settings stored in your repo like what
> is done in this patch.
> This may be a problem for the DPDK github org, where we have DPDK and
> grout starting to use AI bot for reviews, and may have different
> opinions on how to configure the tools.

Well, sure, but I meant in terms of actual usefulness of the tool rather 
than the technicalities of its usage. It may very well be that 
CodeRabbit is more generous and configurable. I was mostly referring to 
Sourcery's ability to summarize review and provide context that, at 
least for the examples that I looked at, seems to be helpful.

> 
> 
>> of an alternate reality where we use GitHub (or something else more
>> modern) to review code :)
> 
> Erm, my personal opinion, the github PR webui is a *mess*.

Just about every review format is going to be a mess, but I'm comparing 
it to patches and patchwork. There's a lot wrong with web UI's such as 
GitHub/GitLab etc. but they do fix *some* issues with patch workflow, 
and there's a lot to be said for having reviewed code right in your IDE 
(which IDE's can do with GitHub integration) where you can debug it and 
provide comments inline. I personally find that to be valuable.

-- 
Thanks,
Anatoly