[ECOS] Ecos on an i386 PC target

[ECOS] Ecos on an i386 PC target

[Paolo Marini]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1821 bytes --]

Hi all,
 
I recently spent some time evaluating Ecos, 
beginning with a PC platform.
 
I downloaded the GNU tools, the Ecos tarball, 
followed the instructions and built a cross development environment for a Linux 
RH62 creating executables for an i386-elf PC. I built the ecos pc stubs tree, 
created the floppy and fired up an old abandoned PC. The GDB (Insight) debugger 
seems to connect to the target (If I disconnect the serial cable, the debugger 
stops for some times).
 
When downloading an executable to the PC target, it 
suddenly resets and reboots as soon as the first packets of code 
arrive.
 
The Linux hosted synthetic i386 target works pretty 
fine (the Insight debugger is quite impressive).
 
Thereafter, I tried with CygWin.
 
I was only able to compile the binutils 
executables; I followed the instructions, unzipped the gcc and gdb (Insight) 
tarballs, applied the specified patches, but the compilation aborted after some 
time. In the case of insight, it seems that it misses some symbols while linking 
the insight executable. The GCC compiles for some time, then stops tossing out a 
lot of errors.
 
Then, I tried the Windows configuration tool for 
ecos. I noticed, for the PC STUBS template, that the COM1 (serial line 1) is 
used for both the debugger and the diagnostic output. I returned to linux, 
regenerated the configuration file, edited it manually by forcing the 
diagnostics to go on the PC screen. This time the target booted up and the 
(still impressive !) insight debugger took control of the target, downloaded the 
executable and run smoothly.
 
Does anyone have some instruction for building the 
CygWin executables for the gcc and gdb ? Is there any new version or patche to 
be applied to the tarball ?
 
Thank you
 
Ciao
 
Paolo 
Marini

Re: [ECOS] Ecos on an i386 PC target

[Jonathan Larmour]

> Paolo Marini wrote:
[ building for i386 pc ]
> Then, I tried the Windows configuration tool for ecos. I noticed, for the
> PC STUBS template, that the COM1 (serial line 1) is used for both the
> debugger and the diagnostic output. I returned to linux, regenerated the
> configuration file, edited it manually by forcing the diagnostics to go
> on the PC screen. This time the target booted up and the (still
> impressive !) insight debugger took control of the target, downloaded the
> executable and run smoothly.

True. This is one of those strange situations where I don't see how this
worked before! Nick, any ideas?
 
> Does anyone have some instruction for building the CygWin executables for
> the gcc and gdb ? Is there any new version or patche to be applied to the
> tarball ?

What errors were you seeing?

Jifl
-- 
Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
"Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault

Re: [ECOS] Ecos on an i386 PC target

[Nick Garnett]

Jonathan Larmour <jlarmour@redhat.com> writes:

> > Paolo Marini wrote:
> [ building for i386 pc ]
> > Then, I tried the Windows configuration tool for ecos. I noticed, for the
> > PC STUBS template, that the COM1 (serial line 1) is used for both the
> > debugger and the diagnostic output. I returned to linux, regenerated the
> > configuration file, edited it manually by forcing the diagnostics to go
> > on the PC screen. This time the target booted up and the (still
> > impressive !) insight debugger took control of the target, downloaded the
> > executable and run smoothly.
> 
> True. This is one of those strange situations where I don't see how this
> worked before! Nick, any ideas?

I am obviously missing something here. Why wouldn't this work? Sending
diagnostics to the debug channel as O packets is the standard
mechanism for nearly all platforms. Or are we talking about emitting
diagnostic messages from the floppy stub program itself?  If this is
the case, where are these coming from? Such messages could very easily
confuse GDB.


-- 
Nick Garnett, eCos Kernel Architect
Red Hat, Cambridge, UK