[En-Nut-Discussion] Arm7 Compiler

Tue Jul 2 19:21:33 CEST 2019

wiegelj writes:
> 
> Hello together. I feel a little bit desperate :-O .
> 
> After a long time I have to look for a network issue, probably in a nut
> application and not in the system. I want to debug with the segger
> j-link and I want to use the EmIde under Windows 10.
> 
> I would like to use the EmIde together with Ethernut 5.1 or Ethernut 
> 4.10.2 both with configfile: at91sam7se-ek.conf
> 
> The 5.1 Version begins to work and it never returns. Also I can't switch
> to the arm-none-eabi compiler. Probaly it can't find a compiler the
> non-abi version is not available in the nut os configurator.
> 
> 17:06:14: Creating Makefiles for arm-gcc in nutbld
> 
> 17:06:14: Creating header files in nutbld
> 
> 17:06:14: ----- Running 'make clean' -----
> 
> 17:06:14: ----- Running 'make all' -----
> 
> 17:06:14: ----- Running 'make install' -----

Well, first attempt should be to manual change to the created directory and
type "make install" there. Output should be more verbose, maybe giving a hint.

> 
> ---------------------
> 
> The version 4.10.2 shows this error/warning:
> arm-none-eabi-ar: creating libnutarch.a
> arm-none-eabi-ar: creating libnutos.a
> arm-none-eabi-ar: creating libnutdev.a
> arm-none-eabi-ar: creating libnutc.a
> arm-none-eabi-ar: creating libnutcrt.a
> ../../nut/gorp/hashes/md5.c: In function 'NutMD5Final':
> ../../nut/gorp/hashes/md5.c:307:5: error: dereferencing type-punned
> pointer will break strict-aliasing rules [-Werror=strict-aliasing]
> ../../nut/gorp/hashes/md5.c:308:5: error: dereferencing type-punned
> pointer will break strict-aliasing rules [-Werror=strict-aliasing]
> cc1.exe: all warnings being treated as errors
> make[1]: *** [hashes/md5.o] Error 1
> make: *** [all] Error 2
> 16:28:27: ----- 'make all' failed with error 2 -----
> 
> Compiler version from the EmIde. I don't know what to do know.
>

It seems you are doing compiler and source archeology. So you should do git/svn archeology.

commit 94a17b0c8833946d2b13cbb396be4ff400ad33aa
Author: u_bonnes <u_bonnes at 1ca430b3-8271-4a6b-9b0e-3dcac98a8924>
Date:   Mon Aug 20 11:30:22 2012 +0000

    Remove no longer needed workaround for aliased variables

    git-svn-id: svn+ssh://svn.code.sf.net/p/ethernut/code/trunk@4467 1ca430b3-8271-
4a6b-9b0e-3dcac98a8924

diff --git a/nut/Makevars.cm3-gcc b/nut/Makevars.cm3-gcc
...
-# Temporarily fixes alignment problems with STM32 CAN driver.
-ifeq ($(PLATFORM), STM32_COMSTICK)
-CFLAGS_OPT += -fno-strict-aliasing
-endif

So at some point no-strict-aliasing was needed. Do you have something related to
aliasing in your options. Perhaps add/remove that option.

Do you need gorp at all? Perhaps try to uncomment the instantiation of gorp in
repository.nut as a dirty workaround.

> Off topic: I think a lot of people switched to Embedded Linux instead of
> nutos? When I'm right, then I don't know what to do with security
> problems. Are you supplying patches to your customer? Or do you
> following a trend to work without an OS? I heard that microchip has a
> stack for TCP-Ip.

I would like to here too.

B.t.w. on SVN head:
qnutconf  at91sam7se-ek.conf ->Build/build
19:00:28: Creating Makefiles for  in /tmp/en
19:00:28: Creating header files in /tmp/en
19:00:28: ----- Running 'make clean' -----
19:00:29: ----- Running 'make all' -----
19:00:34: ----- Running 'make install' -----
19:00:35: Build finished successfully in 6s
->Build/Creat sample
19:00:45: Copying samples from /devel/ethernut/nut/app to /devel/en_apps/gpio/nrf52840_dongle/nrf52840_dongle/nutapp
19:00:45: Creating Makefiles for  in /devel/en_apps/gpio/nrf52840_dongle/nrf52840_dongle/nutapp
19:00:45: OK

Now "cd  /devel/en_apps/gpio/nrf52840_dongle/nrf52840_dongle/nutapp -> make" only shows problems around LED1.

So did you try to build your application against SVN Head?

Bye

-- 
Uwe Bonnes                bon at elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------