[avr-gcc-list] converting gas-generated stabs debugging symbols to Atmel coff-format

[Haase Bjoern (EW-BEU/PMT) *]

Hy,

since this is my first post at this mailing list please allow me one initial 
remark:
I am quite impressed seeing all these activities and the number of people 
working for setting up a free development environment for AVR microcontrollers.

Before I start detailing my problem, some base information:
I am presently trying to convince our department to use GCC as platform for SW 
development for our embedded hardware. Our IT department forces us to use a 
microsoft OS, so I don't see how to avoid using AVR-STUDIO as our debug tool. 
We, thus, need to work with coff-Files :-(. 

Since we will have applications with a certain amount of assembler code, it 
would be *very* helpful if
we could step through our assembler source code during debugging. 
And that is where I have encountered problems: So far, I did not succeed in 
debugging assembler source with the AVRStudio tool. 

(I have scanned the archive since september 03 when the avr-objcopy patch for 
coff-avr has been published. Unfortunately, I did not find a solution therein.)

It seems to me, that the problem is related to the conversion procedure trying 
to translate the ELF- information to the archaic COFF format.

Using the appropriate options, I succeed in including the required debug 
information in the ELF object files,
E.g.:
avr-gcc -Wa,-gstabs -c -mmcu=atmega169 -o lcd.o lcd.s
  ... or ...
avr-as --gstabs -mmcu atmega169 -o lcd.o lcd.s
.
Executing "avr-objdump -S lcd.o" shows that the debugging information is 
generated correctly: It displays the
expected mix between my source code and the disassembly.

It seems, however, that this debugging information is lost when converting the 
object files to coff.
Executing "avr-objcopy -O coff-ext-avr --debugging lcd.o lcd.coff" and 
"avr-objdump -S lcd.coff"
displays only the disassembly and does not include the original source code (as 
does AVRStudio).

Looking forward to hearing from you,

 Björn

******************* address@hidden *******************
Dr. Björn Haase,  Tel.: ++49-711-758 2960
Robert Bosch GmbH, Measurement tools department EW/PMT

P.S.: 
Apart of debugging assembler sources everything seem to work fine :-). I did 
not have any problem so far with
debugging of c-source files and in comparison, (e.g. with Motorola HC08 or NEC 
architectures) the gcc generated
code seems to be fairly compact(!). This might, however, also be related to a 
more efficient controller architecture.

P.P.S.: 
I tried to have a look on the BFD sources written by Jörg Wunsch that implement 
the seemingly 
intricate stabs conversion between the elf and coff formats.

Unfortunately, I do not know where to find recent sources of the required patch 
for the binutils 2.14. Could somebody tell me where to look for them? After 
successfully applying, e.g., all the different patches for the binutils found 
on the web page of Stephan Eisvogel the sources did no longer compile under my 
Cygwin-environment.

P.P.P.S.: 
One question asked by someone who is not *really* familiar with the binutils 
and gcc-source and gnu-based sw development procedures: Is there kind of  
"gentle" introduction to GCC and binutils for somebody who would like to learn 
enough about the project to be able to contribute himself .?