odd error building Stephano's arm9_str91x_eclipse

gorlash3 wrote on Thursday, November 07, 2013:

<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
Okay, I found Stephano's package, which is linked to from the FreeRTOS page, and dates from 2010.
I downloaded his Souce and Common packages, and am trying to build on that. 

Many files have built successfully, until I got to semtest.c, when I got this result:

-------- begin --------
cc (GCC) 4.7.3
Copyright (C) 2012 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

../Common/Minimal/semtest.c -> Debug/semtest.o
cc -c  -ggdb  -Os  -std=gnu89  -TSTR91x-ROM.ld  -I .  -I ./STCode  -I ../Source/include  -I ../Source/portable/GCC/STR9x  -I ../Common/include  -I ../Common/drivers/ST/STR91xFWLib/v2.0/inc  -I ../Common/ethernet/lwIP/v1_3_2/src/include  -I ../Common/ethernet/lwIP/v1_3_2/src/include/ipv4  -I ../Common/ethernet/lwIP/v1_3_2/src/include/lwip  -I ../Common/ethernet/lwIP/v1_3_2/src/include/netif  -I ../Common/ethernet/lwIP/v1_3_2/src/include/ipv4/lwip  -I ../Common/ethernet/lwIP/v1_3_2/contrib/port/FreeRTOS/STR91x  -D STR91X_GCC  -D STACK_LWIP_130  -D NO_THUMB_INTERWORK ../Common/Minimal/semtest.c -o Debug/semtest.o
/cygdrive/c/Users/derelict/AppData/Local/Temp/ccCrl41g.s: Assembler messages:
/cygdrive/c/Users/derelict/AppData/Local/Temp/ccCrl41g.s:152: Error: no such instruction: `swi 0'
makefile:247: recipe for target `Debug/semtest.o' failed
make: *** [Debug/semtest.o] Error 1

//************************************************************
However, I grepped down through the entire source tree (*.c, *.h, *.s), and I don't see "swi 0" present anywhere!!  I was guessing that it was in a macro somewhere, but I don't find it anywhere.  In fact, the only files where I found "swi" at all, are in Microblaze and hcs12 sources, which I am not accessing at all.

Where is this error coming from??




This email is free from viruses and malware because avast! Antivirus protection is active.


rtel wrote on Thursday, November 07, 2013:

Look in portmacro.h, you will probably find something like this:

#define portYIELD() asm volatile ( “SWI 0” )

regards.

gorlash3 wrote on Thursday, November 07, 2013:

Okay, sh#t… I didn’t have “case insensitive” selected when I did my search.

Anyway, I found out what’s really wrong… The makefile in Stephano’s package had a bug:
-D STR91X_GCC
-D STACK_LWIP_130
-D $(THUMB_D)
-mcpu=arm966e-s \

Notice the absence of a backslash after the THUMB_D line - the THUMB_D macro doesn’t contain a backslash either… so the build lines were being truncated at that point!!
Now that I’ve fixed that bug, I’m getting a different error, that -mcpu is deprecated, replaced by -mtune or -march, so I need to find what the equivalent arguments are…

gorlash3 wrote on Thursday, November 07, 2013:

Dammit … now I’m trying to use -march= (as recommended by gnu.gcc.org), but I can’t find any valid argument that this compiler will accept!! The arm966e-s core is an armv5te architecture… gnu.gcc.org says that -march will accept:
‘armv5’, ‘armv5t’, ‘armv5e’, ‘armv5te’,
But I’ve tried all of these, and my (Yagarto) compiler rejects them all!!
How do I proceed from here?? Is there some way to ask gcc which arguments it will accept for -march ??

Later note:
Okay, I figured this one out as well… Even though I was explicitly defining CC to point to my ARM compiler chain, make was actually calling some other compiler; I don’t even know where it’s finding a compiler that is version “cc (GCC) 4.7.3” !!

My three ARM cross compilers are V4.2.2, V4.5.0, and V4.7.4 20130913.

This is very bizarre!! What is clear, however, is that at this point, my problem is with a very complex Makefile, not FreeRTOS.
Thus, I will take this somewhere where I can discuss make !!