Re: RPI: * failed to add service - already in use?

[Michael Blatt <michael.blatt%gmail.com@localhost>]

So the SIGILL is fine, good to know.
It's not dumping core, but failing to open an SDL2 window, hence me
running it through gdb:

lvl2:   EmulationStation - v2.0.0-rc1, built Mar  3 2018 - 19:14:54
lvl2:   Creating surface...
lvl0:   Error creating SDL window!
        Could not initialize OpenGL / GLES library
lvl0:   Renderer failed to initialize!
lvl0:   Window failed to initialize!

retroarch uses SDL1.2 and is able to display its menu.
The raspberrypi-userland samples also compile and run well.

All packages were compiled from 2017Q4 on the same build of 8.0_BETA.
gpu_mem=256 but I have tried other values as low as 32.

Very odd... it is a separate defekt to the libcrypto issues outlined above?

On 6 March 2018 at 08:12, Martin Husemann <martin%duskware.de@localhost> wrote:
> On Tue, Mar 06, 2018 at 07:46:56AM +0000, Michael Blatt wrote:
>> #if __ARM_MAX_ARCH__>=8
>> .global _armv8_aes_probe
>>
>> __ARM_MAX_ARCH should be 6, not 7, since we are on armv6 and do not
>> have neon anyway.
>
> The idea of the cpuid code is to execute the instructions and check
> whether a SIGILL occurs, catch it, and then not use the implementation
> optimized for armv7 or armv8.
>
> This way you get optimal performance even if runing binaries compiled for
> armv6 on armv7 machines.
>
> So the SIGILL itself is perfectly fine, it not being caught and the program
> dumping core, however, is not.
>
> Martin