Re: [dpdk-dev] [PATCH] Make -Werror optional

[Panu Matilainen <pmatilai@redhat.com>]

On 02/21/2015 09:33 PM, Neil Horman wrote:
> On Fri, Feb 20, 2015 at 05:55:21PM -0800, Stephen Hemminger wrote:
>> On Thu, 12 Feb 2015 16:54:44 +0200
>> Panu Matilainen <pmatilai@redhat.com> wrote:
>>
>>> On 02/12/2015 04:38 PM, Stephen Hemminger wrote:
>>>> On Thu, 12 Feb 2015 13:13:22 +0200
>>>> Panu Matilainen <pmatilai@redhat.com> wrote:
>>>>
>>>>> This adds new CONFIG_RTE_ERROR_ON_WARNING config option to enable
>>>>> fail-on-warning compile behavior, defaulting to off.
>>>>>
>>>>> Failing build on warnings is a useful developer tool but its bad
>>>>> for release tarballs which can and do get built with newer
>>>>> compilers than what was used/available during development. Compilers
>>>>> routinely add new warnings so code which built silently with cc X
>>>>> might no longer do so with X+1. This doesn't make the existing code
>>>>> any more buggier and failing the build in this case does not help
>>>>> not help improve code quality of an already released version either.
>>
>> Hopefully distro's like RHEL will build with -Werror enabled
>> and not allow build to go through with errors.
>>
> Thats usually what we do, yes.

Um, nope. All Fedora and RHEL builds are done using a common base set of 
flags set centrally from rpm configuration, and that includes among 
other things -Wall but not -Werror, although since F21 
-Werror=format-security is included since that there are relatively few 
false positives for that.

The thing is, compiler warnings from compilers are just that: warnings, 
and often including hefty dose of false positives. A good package 
maintainer will look at the build logs of his/her packages, investigate 
warnings and send patches upstream to address them in oncoming versions 
where actually relevant, but generally a package maintainer in a distro 
is not responsible for achieving zero-warning build, nor should they.

Take for example set-but-not-used warning introduced in gcc 4.6. Many of 
the warnings unearthed by that do indicate real potential issues in the 
software (a typical example being ignoring an allegedly unlikely error 
code from a syscall, sometimes dozens or even hundreds of them for 
larger projects) but its also typically code that's been in use for 
years and years without anybody noticing. That code does not suddenly 
become unusable just because its now being compiled by a compiler that 
detects this condition, and -Werror on distro level misplaces the burden 
of addressing years of upstream laziness on a packager who often has 
little to do with upstream.

	- Panu -