Re: Fwd: Re: [PATCH AUTOSEL 5.6 082/129] compiler.h: fix error in BUILD_BUG_ON() reporting

[Sasha Levin <sashal@kernel.org>]

On Wed, Apr 22, 2020 at 10:53:33AM +0200, Vegard Nossum wrote:
>
>On 4/22/20 10:34 AM, Greg KH wrote:
>>On Wed, Apr 22, 2020 at 10:21:23AM +0200, Vegard Nossum wrote:
>>>
>>>Hi,
>>>
>>>There is no point in taking this patch on any stable kernel as it's just
>>>improving a build error diagnostic message.
>>
>>build error messages are nice to have be correct, don't you think?
>>
>>Seems like a valid fix for me.
>>
>>thanks,
>>
>>greg k-h
>>
>
>The patch will break on gcc 4.2 and earlier, so if the older stable
>kernels support those then people might be surprised. The mainline patch
>was deemed fine since gcc 4.6 is required. More info here:
><https://lkml.org/lkml/2020/3/31/1477>
>
>If no stable kernel is built with gcc <= 4.2 then you can disregard this.
>
>I think the patch also doesn't really satisfy the following criteria
>from stable_kernel_rules.txt:
>
> - "It must fix a real bug that bothers people": It bothered me (and I
>ran into the bug) on mainline, but that was while writing brand new code
>that used BUILD_BUG_ON(). Presumably these things will not fire on
>existing kernel code.
>
> - "It must fix a problem that causes a build error ...": It doesn't fix
>any of the things listed, not even a build error, just a _diagnostic_
>for an incredibly rare case of a rare kind of build error.
>
>In the end, I am not personally opposed to having the patch in stable,
>but it seems to go against everything I've ever heard about stable rules
>so I'm a bit surprised when you take it anyway. I think it might reduce
>people's trust in stable kernels if any random weird patch is taken
>uncritically when even the patch author points out that it doesn't fit
>the criteria!

Hey Vegard,

I'll drop it.

In general, patches that address build issues are easier to rationalize
in the context of stable as the regressions they might cause will be
limited to build time.

-- 
Thanks,
Sasha