[Buildroot] [autobuild.buildroot.net] Build results for 2014-10-22

[Arnout Vandecappelle <arnout@mind.be>]

On 23/10/14 14:01, Thomas Petazzoni wrote:
> Dear Baruch Siach,
> 
> On Thu, 23 Oct 2014 11:24:16 +0300, Baruch Siach wrote:
> 
>> On Thu, Oct 23, 2014 at 08:30:16AM +0200, Thomas Petazzoni wrote:
>>>          arm |                     ipset-6.23 | NOK | http://autobuild.buildroot.net/results/2ae233d33b05fac10e9b6676a6ca179c75e4c1d9/
>>>     mips64el |                     ipset-6.23 | NOK | http://autobuild.buildroot.net/results/4f55c2415281c6204500efe28fe9e24c8ef73863/
>>
>> I started looking into these build failures and it makes me wonder. The direct 
>> cause of these failures is a combination of two things. Host header directory, 
>> /usr/local/include, being used for target build (clearly a bad thing), and 
>> -Werror. We normally remove -Werror since it is quite useless when you only 
>> want to build the package. The thing is that ipset add -Werror only when 
>> configures with --enable-debug, which in turn is triggered by 
>> BR2_ENABLE_DEBUG. This apparently causes other build issues as well 
>> (http://patchwork.ozlabs.org/patch/400834/). Does it make sense to remove 
>> -Werror in this case? This is after all what the user asks for when enabling 
>> BR2_ENABLE_DEBUG. But then again, does it make sense to have the autobuilder 
>> test BR2_ENABLE_DEBUG, given its relatively high rate of -Werror induced false 
>> positives?
>>
>> What do others think?
> 
> That is indeed a very good question. In some sense, having -Werror
> enabled when --enable-debug is passed is quite legitimate.

 No, I disagree with that. --enable-debug should be an option to enable anything
that makes runtime debugging easier - sometimes it enables extra debugging
output for instance.  -Werror is something which is only really valid for
continuous integration.



> On the other
> hand, the reason I'm testing BR2_ENABLE_DEBUG=y in the autobuilders is
> because users could do it, so I prefer to have this case tested as
> well. My reasoning is that if we have an option in Buildroot, then it
> should be tested and working. Otherwise, we should get rid of the
> option.

 +1 to that.

> 
> Maybe the fact that BR2_ENABLE_DEBUG=y passes --enable-debug is wrong?
> Maybe BR2_ENABLE_DEBUG=y should really only build all packages with
> debugging symbols (passing -g to gcc), and no try to enable per-package
> debugging features?

 I disagree with this one as well. As I said above, there's a lot more that can
be enabled with --enable-debug than just a -g.


 Regards,
 Arnout

> 
> Any input from others? Arnout, Peter?
> 
> Thomas
> 


-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F