Re: aarch64 crash with static PIE

[Carlos O'Donell <carlos at redhat dot com>]

On 01/19/2018 12:40 PM, Szabolcs Nagy wrote:
> * Carlos O'Donell <carlos@redhat.com> [2018-01-19 12:09:53 -0800]:
>> On 01/19/2018 11:48 AM, Florian Weimer wrote:
>>> On 01/19/2018 08:39 PM, H.J. Lu wrote:
>>>>> Not sure what is going on.  This is with:
>>>>>
>>>>> binutils-2.29.1-12.fc28.aarch64
>>>
>>>> Don't you need binutils 2.30?
>>>
>>> Our installation instructions do not mention a specific binutils
>>> requirement.
>> This was an oversight on my part. We should document this in NEWS
>> along with the statement of support. It will be different from the
>> INSTALL requirement for building the project with default configure
>> options. We could extend configure tests to cover this, but I think
>> it's excessive since a given version of binutils is not what matters,
>> but instead the individual capabilities that need bug fixes which
>> may be backported by vendors.
>>
> 
> that's why i was hesitant to update the news..
> https://sourceware.org/ml/libc-alpha/2017-12/msg00555.html
> 
> it currently says gcc-8 and does not mention binutils,
> even on x86 there is a minimum binutils requirement
> for this to work, on aarch64 the requirement is just
> higher: at least the fixes for ld/22263 and ld/22269
> are required which are in 2.30 now.
> 
> i didnt write configure tests for the bugs thinking
> this is an experimental feature, may be i should have..

The feature deserves a NEWS entry.

We should say binutils 2.30 is required.

We should strive to have configure checks for all things that could
go wrong with *default* configure options.

For non-default configure options you have to know what you are
doing and understand the consequences. Once we stabilize the
feature a bit more we can *then* add configure checks IMO.

In summary: Can you please add a binutils 2.30 requirement to the
NEWS entry?

-- 
Cheers,
Carlos.