key steps to get c-file debugging info into AVRStudio 4

What are the key steps to getting AVRStudio 4.11 to show C source file
debugging info (rather than an assembly listing of the flash memory)?

I've been trying different things for hours and nothing seems to work.

I have an ATmega8535.

Is there any documentation that focuses on the 20% of things that 80% of
users would like to do (I'm assuming this to be something like the
following):

Development Environment:
    Native Windows (2k or XP) - (not Linux, Unix or Cygwin).

Development Language:
    C

Deep knowledge of compiler, object, linker internals, etc.:
    Zero.

Number of Files in project:
    More than one, and more functions than main() alone.

Thanks.

Joe wrote:

>What are the key steps to getting AVRStudio 4.11 to show C source file
>debugging info (rather than an assembly listing of the flash memory)?
>
>  
>
What compiler? If GNU C / WinAVR, can't find ya. Check out 
www.avrfreaks.net for info on that. If ImageCraft's ICCAVR, then just 
build the project under ICC, then open the COFF .cof file in Studio. 
Viola, source code debugging.

-- 
// richard
http://www.imagecraft.com

yes, it is gcc that I'm using.

I've been searching avrfreaks, but it's no help - a lot of the links are now
broken, and anyway, the elf2coff thing, I already have - but it doesn't seem
to work with the latest gcc ("can't create elf dbase").

I did load ImageCraft's ICCAVR with a view to porting the existing project
to this environment, but I got put off by two things:

1.  The keyword that you use to indicate Program Memory (flash), is 'const'.
I got to that while porting and thought it seemed suspect.  I.e. I didn't
imagine that it would always do what I was wanting.

2. When I was trying to convert an old gcc 'set bit' instruction, your
documentation said, "see the Atmel datasheet".  I'm afraid that was the
deciding factor for me - to stick with GCC for a bit longer to see if I
could get it to work.

When one pays for a product, one does not want to be sent running around
looking for answers - if a runaround IS necessary, at least and explanation
of why and a hint of where in the datasheet to look would be helpful.

3. (This was not a deciding factor, but it's a negative factor with your
product)  Whenever I open your html help, I keep getting asked if I want to
allow such and such a thing to be installed (I don't, so this is very
annoying) - what does this give me that I can't do without it- I seem to be
able to browse your help even if it does not load.

What are your comments on these points?

I'm in no way a commited gcc person and would willingly pay to make my life
easier when it comes to developing this code, so there is still time to
convert me to ICCAVR.

Note, this was an inherited project that was developed with the AT90S8535
(but it has to be moved over to the ATmega8535) - hence the change to the
latest gcc (the gcc it was originally developed with didn't seem to support
the mega version, nor did version 3 of AVRStudio).  The migration is
complete except for the c-level debugging info.

Joe



"Richard M." <richard@imagecraft.com> wrote in message
news:113d3j8j31s5c44@corp.supernews.com...
> Joe wrote:
>
> >What are the key steps to getting AVRStudio 4.11 to show C source file
> >debugging info (rather than an assembly listing of the flash memory)?
> >
> >
> >
> What compiler? If GNU C / WinAVR, can't find ya. Check out
> www.avrfreaks.net for info on that. If ImageCraft's ICCAVR, then just
> build the project under ICC, then open the COFF .cof file in Studio.
> Viola, source code debugging.
>
> --
> // richard
> http://www.imagecraft.com

On Tue, 15 Mar 2005 12:01:46 -0000, "Joe"
<ffffh.no.spam@hotmail-spammers-paradise.com> wrote:

>the mega version, nor did version 3 of AVRStudio).  The migration is
>complete except for the c-level debugging info.
>
>Joe
>

Studio 4.11 uses elf-dwarf2 for debugging info.

I think i was somewhere that you sould replace the -g <blank> option 
 with -g elf-dwarf 

Check if elf-dwarf is correct spelled or if its called 	elf-dwarf2

Load the elf file into Studio & Roll ...

carsten

I would embrace you and give you a guy-to-guy hug if I could but reach out.

That's done it!

The exact thing (out of all the permutations that I had, unsuccessfully,
tried last night was):

As you said, use: -gdwarf-2
(elf-dwarf2 wasn't recognised by the compiler)

This is it, in its full context (the local make file - not including the
avrfreaks/avr_make additions that get added to the compiler flags when it is
included.

#compiler flags - optimise for space
 CPFLAGS
= -gdwarf-2 -Os -Wall -w -Wstrict-prototypes -Wa,-ahlms=$(<:.c=.lst)

Next important thing:
Open the <target>.elf file produced by the build (not the <target>.obj
file), and that's done it (c source files now run thru AVRStudio 4.11).

Thanks a lot.

Some other stuff that may help others if they use gcc 3.4.3 (from
WinAVR-20050214-install.exe):
<don't assume that I'm right here, it's just what I had to do to get a
build>

In the avrfreaks/avr_make file, I had to change block 8 because avr-objcopy
does not seem to support the avrobj format anymore.

#%.obj: %.elf
# $(BIN) -O avrobj -R .eeprom $< $@

%.obj: %.elf
 $(BIN) -O coff-ext-avr -R .eeprom $< $@


also modified the following in block 9 (this is on a Win2k system - don't
know if it's still valid for Linux):
%.cof: %.elf
 $(ELFCOF) $< $(OUT) $@ $*.sym
 $(CP) $(OUT)/$@ .
 $(CP) $(OUT)/*.sym .
 $(CP) $(OUT)/*.cof .
# $(CP) $(OUT)/* .
# $(CP) $(OUT)\\*sym .
# $(CP) $(OUT)\\*S .

I'm still getting elfcoff saying that it was, "Unable to create Elf dbase",
but everything else seems okay now.



"Carsten" <xnews1@luna.kyed.com> wrote in message
news:uk6e31pvg0196ubbgru0b0i0arf62dt75k@4ax.com...
> On Tue, 15 Mar 2005 12:01:46 -0000, "Joe"
> <ffffh.no.spam@hotmail-spammers-paradise.com> wrote:
>
> >the mega version, nor did version 3 of AVRStudio).  The migration is
> >complete except for the c-level debugging info.
> >
> >Joe
> >
>
> Studio 4.11 uses elf-dwarf2 for debugging info.
>
> I think i was somewhere that you sould replace the -g <blank> option
>  with -g elf-dwarf
>
> Check if elf-dwarf is correct spelled or if its called elf-dwarf2
>
> Load the elf file into Studio & Roll ...
>
> carsten
>

Joe wrote:

>yes, it is gcc that I'm using.
>...
>I did load ImageCraft's ICCAVR with a view to porting the existing project
>to this environment, but I got put off by two things:
>
>1.  The keyword that you use to indicate Program Memory (flash), is 'const'.
>I got to that while porting and thought it seemed suspect.  I.e. I didn't
>imagine that it would always do what I was wanting.
>
>  
>
While somewhat unusual, I don't think it breaks ANSI C's definition per 
se, except for people who use "const char *" to a function's argument to 
mean the function will not change the content, but I can be wrong. In 
anyway, AVR is a Harvard architecture so this issue needs to be handled 
one way or another...

I *may* add a separate keyword at some future release, but it will have 
to be done slowly to support backward compatibility :-(

>2. When I was trying to convert an old gcc 'set bit' instruction, your
>documentation said, "see the Atmel datasheet".  I'm afraid that was the
>deciding factor for me - to stick with GCC for a bit longer to see if I
>could get it to work.
>
>  
>
Eh? One problem here is that the old GCC does not generate set bit 
instruction so now you have unnecessary extensions to the language in 
your program. I am pretty sure the new GCC fixes it, but a set bit is 
simply:

unsigned char c;
#define bit(x)   (1 << (x))
    c |= bit(2);
...
etc.

>When one pays for a product, one does not want to be sent running around
>looking for answers - if a runaround IS necessary, at least and explanation
>of why and a hint of where in the datasheet to look would be helpful.
>
>  
>
Unless I am wrong, you are only experiencing this because your program 
happens to use a non-standard C extensions. This is the prime reason we 
avoid extensions as much as possible....

>3. (This was not a deciding factor, but it's a negative factor with your
>product)  Whenever I open your html help, I keep getting asked if I want to
>allow such and such a thing to be installed (I don't, so this is very
>annoying) - what does this give me that I can't do without it- I seem to be
>able to browse your help even if it does not load.
>
>  
>
Not sure what you mean here. We switched to use HTML Webwork help on V7, 
so there may be requirements that I don't know of, but so far, I have 
not gotten any such complaints from other people. You need either 
Javascript or Java and the Webworks suppose to detect that, but I am not 
aware that it needs anything else to be installed.

>What are your comments on these points?
>
>I'm in no way a commited gcc person and would willingly pay to make my life
>easier when it comes to developing this code, so there is still time to
>convert me to ICCAVR.
>
>  
>
Well, hope this helps. If you wish, you can email me privately if you 
have more questions etc.

-- 
// richard
http://www.imagecraft.com

Thanks Richard,

I had given up on gcc by the time I read your first post and set about
porting to ICCAVR again - this was fairly easy after all (with a few macros
defined in a new header).  After the first file compiled without warning, I
found Carsten's suggestion on the gcc fix (which worked) - so halted the
port - it's not so important now (given that I recall the demo version of
ICCAVR has a 4k code limit - did I recall that correctly) and I know that
the gcc code so far is 6k (this may become a factor in the abandonment of
gcc).

Anyway, this is what pops up every time I open ICCAVR's help file:
Do you want to install and run: file://C:\iccv7avr\Help\ICCAVR\wwhelp3.cab
(Quadralay Corporation).  As I say, I'd be interested in what this gives me,
since the help seems to work fine when I say "No, I don't want to install
this".

There is a bad downside, actually.  When I have the ICCAVR help open, I get
errors when I open my normal IE browser (to google's home page or any new
website):

"A runtime error has occurred.  Do you wish to debug?
Line: 0
Error: object doesn't support this property or method" - where I have to
click no.

(I am using a fairly old popup killer - proxomitron - maybe this is a
factor) - anyway, not nice.
IE5.5 SP2, Win2k.

A couple of other things that are putting me off the product:

It is not possible to put two source windows side by side in the IDE (to
view one while typing into another).

There is no pre-configured option to use MS Dev Studio's editor as an
alternative (maybe it can't be used? maybe I'm missing the point - but it is
a good editor that has never caused me any problems.).

It's too fussy using the current menu setup to compile a single file:
(especially when one is in an edit and compile cycle to try and figure
something awkward).  MS Dev Studio gets it right when it puts build and
compile under the same 'build' menu (Alt, B, B or Alt, B, C).

Closing a window is too fussy as well - It's a shame you didn't do an MDI
application but chose the tabs (which are good - but you really need the
other things associated with MDI apps to turn them into an asset rather than
simply a replacement).

These are only my personal opinions, and as I've metioned, I've not ruled
out a switch to ICCAVR yet - it will become a necessity if the final code
won't fit (but then, I'll be considering IAR Workbench under this scenario).

Thanks.

Joe


"Richard M." <richard@imagecraft.com> wrote in message
news:113er91out06mc1@corp.supernews.com...
> Joe wrote:
>
> >yes, it is gcc that I'm using.
> >...
> >I did load ImageCraft's ICCAVR with a view to porting the existing
project
> >to this environment, but I got put off by two things:
> >
> >1.  The keyword that you use to indicate Program Memory (flash), is
'const'.
> >I got to that while porting and thought it seemed suspect.  I.e. I didn't
> >imagine that it would always do what I was wanting.
> >
> >
> >
> While somewhat unusual, I don't think it breaks ANSI C's definition per
> se, except for people who use "const char *" to a function's argument to
> mean the function will not change the content, but I can be wrong. In
> anyway, AVR is a Harvard architecture so this issue needs to be handled
> one way or another...
>
> I *may* add a separate keyword at some future release, but it will have
> to be done slowly to support backward compatibility :-(
>
> >2. When I was trying to convert an old gcc 'set bit' instruction, your
> >documentation said, "see the Atmel datasheet".  I'm afraid that was the
> >deciding factor for me - to stick with GCC for a bit longer to see if I
> >could get it to work.
> >
> >
> >
> Eh? One problem here is that the old GCC does not generate set bit
> instruction so now you have unnecessary extensions to the language in
> your program. I am pretty sure the new GCC fixes it, but a set bit is
> simply:
>
> unsigned char c;
> #define bit(x)   (1 << (x))
>     c |= bit(2);
> ...
> etc.
>
> >When one pays for a product, one does not want to be sent running around
> >looking for answers - if a runaround IS necessary, at least and
explanation
> >of why and a hint of where in the datasheet to look would be helpful.
> >
> >
> >
> Unless I am wrong, you are only experiencing this because your program
> happens to use a non-standard C extensions. This is the prime reason we
> avoid extensions as much as possible....
>
> >3. (This was not a deciding factor, but it's a negative factor with your
> >product)  Whenever I open your html help, I keep getting asked if I want
to
> >allow such and such a thing to be installed (I don't, so this is very
> >annoying) - what does this give me that I can't do without it- I seem to
be
> >able to browse your help even if it does not load.
> >
> >
> >
> Not sure what you mean here. We switched to use HTML Webwork help on V7,
> so there may be requirements that I don't know of, but so far, I have
> not gotten any such complaints from other people. You need either
> Javascript or Java and the Webworks suppose to detect that, but I am not
> aware that it needs anything else to be installed.
>
> >What are your comments on these points?
> >
> >I'm in no way a commited gcc person and would willingly pay to make my
life
> >easier when it comes to developing this code, so there is still time to
> >convert me to ICCAVR.
> >
> >
> >
> Well, hope this helps. If you wish, you can email me privately if you
> have more questions etc.
>
> --
> // richard
> http://www.imagecraft.com

Joe wrote:

>Thanks Richard,
>
>I had given up on gcc by the time I read your first post and set about
>porting to ICCAVR again - this was fairly easy after all (with a few macros
>defined in a new header).  After the first file compiled without warning, I
>found Carsten's suggestion on the gcc fix (which worked) - so halted the
>port - it's not so important now (given that I recall the demo version of
>ICCAVR has a 4k code limit - did I recall that correctly) and I know that
>  
>
It's fully functional for 45 days, and then 4K limited "forever" for non 
commercial sue.

>the gcc code so far is 6k (this may become a factor in the abandonment of
>gcc).
>
>Anyway, this is what pops up every time I open ICCAVR's help file:
>Do you want to install and run: file://C:\iccv7avr\Help\ICCAVR\wwhelp3.cab
>(Quadralay Corporation).  As I say, I'd be interested in what this gives me,
>since the help seems to work fine when I say "No, I don't want to install
>this".
>  
>
Hmm... no idea. What OS are you using, and what is your default browswer 
(and version). Do you have Javascript or Java enabled? Do you get to see 
any of the Help at all?

>There is a bad downside, actually.  When I have the ICCAVR help open, I get
>errors when I open my normal IE browser (to google's home page or any new
>website):
>
>"A runtime error has occurred.  Do you wish to debug?
>Line: 0
>Error: object doesn't support this property or method" - where I have to
>click no.
>
>(I am using a fairly old popup killer - proxomitron - maybe this is a
>factor) - anyway, not nice.
>IE5.5 SP2, Win2k.
>  
>
I guess this is the problem using a (well known 3 rd party) product. I 
have no idea... I will take your info and send it to Qudralay. I do 
again stress that no one else has the problems (so far) so it's probably 
something about your particular setup.

>A couple of other things that are putting me off the product:
>
>It is not possible to put two source windows side by side in the IDE (to
>view one while typing into another).
>
>  
>
Not currently. Sorry.

>There is no pre-configured option to use MS Dev Studio's editor as an
>alternative (maybe it can't be used? maybe I'm missing the point - but it is
>a good editor that has never caused me any problems.).
>
>  
>
Yes, you can use any external editor. See Tools->Environments and set 
the editor to external.

>It's too fussy using the current menu setup to compile a single file:
>(especially when one is in an edit and compile cycle to try and figure
>something awkward).  MS Dev Studio gets it right when it puts build and
>compile under the same 'build' menu (Alt, B, B or Alt, B, C).
>
>  
>
Generally we don't recommend people to compile a single file. Just build 
the whole project. This is one of those ease of use thing. For 99% of 
people, they are happy with just clicking on the build icon. If we make 
"compile to single file" easy to use, for some reasons, then lots of 
people start to use it and wonder why they can't build executables etc. 
By moving it away, we solve a lot of support issues w/o hurting most 
users. Just right now, we have a customer wondering why the "linker 
file" blinks and disappears when he clicks on it. It took me many emails 
to fish it out of him that he builds and project, compile each file to 
object, then click on the linkw.exe icon (and expect what?...) Since 
ilinkw is a command line program, of course it just blinks and 
disappears. Some people have very clear idea on what they are doing. 
Unfortunately, they could be wrong at times :-) Our philosophy is to 
make our tools as simple to use as possible for the 80-90% of the users, 
and give the rest some of the hooks.

>Closing a window is too fussy as well - It's a shame you didn't do an MDI
>application but chose the tabs (which are good - but you really need the
>other things associated with MDI apps to turn them into an asset rather than
>simply a replacement).
>
>  
>
We use to have MDI on V5, but again, a whole different set of support 
problems cropped up. Since most people have their own favorite editor 
anyway, we just make our editor usable and spend the energy on the 
compiler and the application builder.
BTW, ^W closes an editor window.


-- 
// richard
http://www.imagecraft.com

On Tue, 15 Mar 2005 17:25:43 -0800, "Richard M."
<richard@imagecraft.com> wrote:

>Joe wrote:
[snip...snip...]
>>There is a bad downside, actually.  When I have the ICCAVR help open, I get
>>errors when I open my normal IE browser (to google's home page or any new
>>website):
>>
>>"A runtime error has occurred.  Do you wish to debug?
>>Line: 0
>>Error: object doesn't support this property or method" - where I have to
>>click no.
>>
>>(I am using a fairly old popup killer - proxomitron - maybe this is a
>>factor) - anyway, not nice.
>>IE5.5 SP2, Win2k.
>>  
>>
>I guess this is the problem using a (well known 3 rd party) product. I 
>have no idea... I will take your info and send it to Qudralay. I do 
>again stress that no one else has the problems (so far) so it's probably 
>something about your particular setup.

FWIW, I have no problems with the ICC AVR help being open in Opera and
then launching another tab on, say, Google.

Although I almost never use IE, pasting the path to the ICC AVR help
file works OK with MSIE, although I do get an IE warning about "Internet
Explorer has restricted this file from showing active content that could
access your computer." 

I guess that MS has so many security holes that when they decide to
block an active control in IE, they need to be all showy about it to
Demonstrate That They Are Taking Security Seriously and are Doing
Something About It. 

No problems, tho, with firing up other IE sessions.

-- 
Rich Webb   Norfolk, VA

"Richard M." <richard@imagecraft.com> wrote in message
news:113er91out06mc1@corp.supernews.com...
> Not sure what you mean here. We switched to use HTML Webwork help on V7,

Mmm..... and now it takes 6 seconds on my 2.4GHz machine to show the first
help page, as opposed to less than 1 second in V6.... But apart from that: a
great compiler and IDE! :-)

Meindert