Re: [PATCH 03/12] Mark all machinery needed in early static-link init as -fno-stack-protector.

[Nick Alcock <nick dot alcock at oracle dot com>]

On 15 Dec 2016, Florian Weimer outgrape:

> On 12/15/2016 03:21 PM, Nick Alcock wrote:
>> On 15 Dec 2016, Florian Weimer outgrape:
>>
>>> On 11/28/2016 01:32 PM, Nix wrote:
>>>> diff --git a/csu/Makefile b/csu/Makefile
>>>> index 31e8bb9..22afe67 100644
>>>> --- a/csu/Makefile
>>>> +++ b/csu/Makefile
>>>> @@ -45,6 +45,11 @@ before-compile += $(objpfx)version-info.h
>>>>  tests := tst-empty tst-atomic tst-atomic-long
>>>>  tests-static := tst-empty
>>>>
>>>> +CFLAGS-.o += $(no-stack-protector)
>>>
>>> This also applies to the tests in the csu subdirectory, which is probably not what we want.
>>
>> Definitely not!
>>
>>> We have two options here: List the .c source files explicitly (like you do in string/), or move the test to a different subdirectory
>>> (perhaps misc/ or stdlib/).  In the second case, there should be an empty “tests” variable assignment in csu/Makefile which
>>> documents why there can't be any tests in this subdirectory.
>>
>> Listing explicitly seems likely to be horrifically hard to maintain. I'm
>> not even sure how many .o's there *are* that this would apply to.
>
> I think you added some magic to rtld CFLAGS to get this working without eval.

My argument just collapsed because I was looking at elf/, not csu/. csu/
is comparatively tiny and much easier to fix things up for. :)

> So maybe moving the tests is the best option here.

Yeah. None of these tests actually look like they belong in csu/, except
maybe tst-empty.  misc/ seems like a better place for all of them.

I'll whip up a moving patch (it'll be a bit longer because I think I
should actually do at least one quick test run before sending it).

-- 
NULL && (void)