commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gilles <>
Subject Re: [ALL] Automated requirements (e.g. CheckStyle)?
Date Tue, 08 Aug 2017 23:13:43 GMT

On Wed, 9 Aug 2017 00:20:00 +0200, Karl-Philipp Richter wrote:
> Hi,
> Am 07.08.2017 um 15:09 schrieb Gilles:
>> Less work for the maintainers is good. :-)
>> By "taking time" I meant that validating should not be enforced when
>> calling "mvn compile" or "mvn test".
> I wouldn't worry about the time consumption of the validation even if
> it's run by every dev before very compilation since the sum of
> conversation parts in patch/PR discussion in the form of "LGTM, 
> except
> for the indentation at line xy" - "OK, I fixed that now" - "Oh, no 
> wait,
> I forgot the trailing space at line yz" - ... - merged takes an 
> infinite
> more of time and energy.

I agree, but that is with respect to interaction with someone not used
to the coding style/rules; what I meant is that when doing one's "own"
work, one shouldn't have to wait for CheckStyle at every compilation,
when you know that you'll fix the missing doc _after_ fixing the code.

> Regarding the phase where checkstyle should be run I have some
> additional thoughts to my initial post:
>   * If running checkstyle will be enforced the only phase that makes
> sense is `validate` because you don't won't to build something that's
> invalid because it's somehow unlogical and a waiste of time if you 
> don't
> fail the build as early as possible. In order to avoid annoyance for
> users who aren't used to fix checkstyle errors before being able to
> build I'd suggest a profile with deactivated checkstyle which allows
> that rather an putting checkstyle in a separate profile.

IIUC, that would be fine (since it takes care of the above scenario).

>   * Running checkstyle in the site or any other reporting phase is in
> Maven speak afaik "show what might be wrong with my build given the 
> fact
> that I consider it passing after compilation, unit and integration 
> tests
> passed" or "show me some statistics about style issues - 150, wow 
> that's
> 12 less than last build".

That's what we've done up to now; and the number of errors is supposed
to be zero before a release.  But I agree that the risk of a lot of 
for the RM would be reduced by enforcing checks at least before 
to the "master" branch.

Is anyone objecting?

I think that the profile should be defined in the "parent" POM.
Can someone make the necessary additions?


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message