Re: [avr-gcc-list] Program Listing With C Comments

[Vincent Trouilliez]

On Mon, 2006-04-03 at 12:44 -0500, David Kelly wrote:
> Vincent, thats originally my Makefile and object.list should already  
> contain exactly what you are looking for.

Hi David ! Yes, I still use your Makefile, I quite like it :-)
It makes it very easy to add more files to my project, and generates all
the files I might need...


> Change the above to -DS. I've built your sources and verified. Don't  
> know when the "S" disappeared as its still in my version.

I removed it today to see if it would make any difference, since someone
suggested this flag on the list. I also tried -S and notice a big
improvement over -DS, in that, with -S, the bags of strings I have
defined, appear like this :

0000009c <menu_main>:
      9c:       05 5f 5f 5f 5f 5f 5f 5f 20 4d 61 69 6e 20 5f 5f     ._______ 
Main __
      ac:       5f 5f 5f 5f 5f 00 45 72 72 6f 72 20 63 6f 64 65     _____.Error 
code
      bc:       73 20 20 20 20 20 20 20 00 00 00 91 02 50 61 72     s       
.....Par
      ...................
      ...................

instead of miles of meaningless assembly "instructions".
It makes much more sense, more useful, and is also a lot more compact if 
nothing else...

With regard to the C/assembly mix problem though, it doesn't change a
thing it seems (no better no worse), so I might as well keep using -S
instead of -DS, unless something important escaped me of course.

> Speaking of having built your sources, 3.4.6 complains on "gmake  
> depend":
> 
> In file included from ui.h:11,
>                   from menu.h:7,
>                   from main.c:14:
> /usr/local/lib/gcc/avr/3.4.6/../../../../avr/include/avr/eeprom.h: 
> 172:3: #error "No write enable bit known for this device's EEPROM."

Hmmm, I have 3.4.3 and it doesn't complain about this... though it does
complain about many other things !! ;-) but it compiles anyway.

>OK, I spoke too soon. For example ui_message_show source is totally  
>separate from its disassembly listing. Get lots of warnings like this:

>ui.c:762: warning: passing arg 1 of `eeprom_read_byte' makes pointer  
>from integer without a cast

Ah yes, I remember trying to get rid of these specific warnings, but I
didn't see the problem, as I was using the exact syntax found in
avr-libc's documentation. I think I fixed it with some type casts, but
that made the eeprom_read_byte() very difficult to read :-(
The documentation says that the address can be passed as a simple
integer, but gcc is less arranging and wants a pointer not an integer...
Well, I will fix it again and see if that improves the output of
objdump...


>...which might contribut to avr-gcc getting confused and not properly  
> tagging source with code.

Hmmm, thanks. Makes sense indeed. objdump can only do as good as the elf
gcc passes on to it...
I will try fixing all my warnings then. The problem is that I think some
of them I never managed to fix, like:

menu.c:53: warning: initialization from incompatible pointer type
menu.c:67: warning: initialization discards qualifiers from pointer
target type


Regards,


--
Vince