[dpdk-dev] [RFC] devtools: improve DPDK review process

[Burakov, Anatoly]

On 01-Apr-21 10:50 AM, Gaëtan Rivet wrote:
> On Thu, Apr 1, 2021, at 11:05, Anatoly Burakov wrote:
>> A common problem on the DPDK mailing list is lack of reviewers for
>> patches. This script is intended to help with that problem, by picking
>> a random commit author from our git history, and adding a Reviewed-by:
>> tag. This can be added to e.g. commit-msg git hook, like so:
>>
>>      # add reviewer to commit message
>>      $(git rev-parse --show-toplevel)/devtools/get-reviewer.sh >> $1
>>
>> Now all DPDK code will definitely be well reviewed!
>>
>> Signed-off-by: Anatoly Burakov <anatoly.burakov at intel.com>
>> Rеviеwеd-by: Thоmas Mоnjаlоn <thomas at monjalon.net>
>> ---
>>   devtools/get-reviewer.sh | 4 ++++
>>   1 file changed, 4 insertions(+)
>>   create mode 100755 devtools/get-reviewer.sh
>>
>> diff --git a/devtools/get-reviewer.sh b/devtools/get-reviewer.sh
>> new file mode 100755
>> index 0000000000..f4b07c9862
>> --- /dev/null
>> +++ b/devtools/get-reviewer.sh
>> @@ -0,0 +1,4 @@
>> +#!/bin/sh
>> +
>> +# find a reviewer for the code
>> +git log --format="Reviewed-by: %an <%ae>" | sort -u | shuf -n 1
>> -- 
>> 2.25.1
>>
>>
> 
> I think this is great.
> However if you accelerate only one end of the pipe, you won't gain much.
> 
> Humans are still required to write patches, this is inefficient.
> Here is a suggestion:
> 
> diff --git a/devtools/infinite-monkey.sh b/devtools/infinite-monkey.sh
> new file mode 100755
> index 0000000000..7d637266d3
> --- /dev/null
> +++ b/devtools/infinite-monkey.sh
> @@ -0,0 +1,13 @@
> +#!/usr/bin/env sh
> +
> +command -v csmith > /dev/null || {
> +    echo 'Require csmith: https://github.com/csmith-project/csmith'
> +    exit 1
> +}
> +
> +until ninja -C build; do csmith > $(find . -name '*.c' |shuf -n 1); done
> +
> +git add -F .
> +git commit -sm "$(dirname file): $(($(date +%s%N)/1000000))" \
> +           -sm '' -sm "$(./devtools/get-reviewer.sh)"
> +git send-email -1 --thread --to dev at dpdk.org
> 
> With proper perf testing at the other end, DPDK should be able to finally become web-scale.
> 

Yep, should be doable to achieve line rate for patches to DPDK!

-- 
Thanks,
Anatoly