Review guidelines and rpmlint

[Hans de Goede]

Patrice Dumas wrote:
> On Sun, Mar 11, 2007 at 02:22:16PM -0500, Callum Lerwick wrote:
>> If you look closely, rpmlint output is separated into warnings (W:) and
>> errors (E:). Just like a C compiler.
> 
> If I remember well Ville said that the difference between W and E was
> quite arbitrary. 
> 
>> Errors MUST be fixed to pass review. (Or to pass the upcoming
>> rpmlint-after-build test)
> 
> Some errors have to be ignored (from the top of my head, errors about
> setuid binaries, for example).
> 
>> Warnings can be ignored IF (BIG IF) there is reasonable justification.
>> For example, the common "no documentation" warning with sub-packages
>> such as -devel. Often all documentation goes into the main package, and
>> there's nothing suited to go in the -devel package.
> 
> In my opinion it should be like that for W and E rpmlint messages 
> indistinctly. Maybe E may be scrutated more, but it isn't obvious either.
> 

Agreed, as a loyal rpmlint user and someone with quite a few packages 
and reviews on his name, I must say this is the only way. Many rpmlint 
errors are errors in most cases but not always, some should / could be 
changed to warnings. But there will never be a rpmlint without false 
positives on the error front.

This is also why my suggestion for adding rpmlint to the buildsys and 
makefile.common test builds, contains a whitelist.

Regards,

Hans