Fwd: All mips BSPs cannot link psxfenv01

[Joel Sherrill]

I'm forwarding a discussion between Sebastian Huber and myself. Looks like
the mips fenv.h support was broken recently.

Any ideas?

---------- Forwarded message ---------
From: Sebastian Huber <sebastian.huber@embedded-brains.de>
Date: Sat, Jan 15, 2022, 1:01 PM
Subject: Re: All mips BSPs cannot link psxfenv01
To: <joel@rtems.org>
Cc: RTEMS <devel@rtems.org>


On 14/01/2022 17:51, Joel Sherrill wrote:
>> Is this due to the recent Newlib build system changes?
> I guess so although I don't see anything obvious when I
> looked at the machine/configure.ac in both versions.
>
>   I confirmed jmr3904 could build all tests with this gcc/newlib:
>
> $ ~/rtems-work/tools/6/bin/mips-rtems6-gcc --version
> mips-rtems6-gcc (GCC) 10.3.1 20211214 (RTEMS 6, RSB
> bb9e049b103f3a47019a6cfad18e2b59c9e9ea97, Newlib eb03ac1)
>
> That should have been the immediately previous version.

It is probably this change:

commit 8fa73a9f8414a4926365324c2fe32a237c2eb91d
Author: Jeff Johnston <jjohnstn@redhat.com>
Date:   Mon Nov 29 17:33:52 2021 -0500

     Modifying patch from: marian.buschsieweke@ovgu.de

     The code accessing the floating point control/status register, namely

             #define __cfc1(__fcsr)  __asm __volatile("cfc1 %0, $31" :
"=r" (__fcsr)

     does not compile with mips16. This changed the makefile to pass
-mno-mips16 to avoid the following
     compiler error:

             mips-mti-elf fails with "Error: unrecognized opcode `cfc1
$3,$31'"

I have no time to dig into the Newlib build system.

-- 
embedded brains GmbH
Herr Sebastian HUBER
Dornierstr. 4
82178 Puchheim
Germany
email: sebastian.huber@embedded-brains.de
phone: +49-89-18 94 741 - 16
fax:   +49-89-18 94 741 - 08

Registergericht: Amtsgericht München
Registernummer: HRB 157899
Vertretungsberechtigte Geschäftsführer: Peter Rasmussen, Thomas Dörfler
Unsere Datenschutzerklärung finden Sie hier:
https://embedded-brains.de/datenschutzerklaerung/