[ECOS] Problem downloading eCos

[ECOS] Problem downloading eCos

[Daniel.Andersson]

Hi,

I have just recently got my EB40 board on-line with gdb (using
serial/angel). However, i cannot download anything to the memory. I can view
the memory through insight 5.0 but when i connect to target and issue a
download the whole environment hangs. Has it anything to do with my ecos
configuration or is it something else?

Regards, Daniel

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

Daniel.Andersson@combitechsystems.com wrote:
> 
> Hi,
> 
> I have just recently got my EB40 board on-line with gdb (using
> serial/angel). However, i cannot download anything to the memory. I can view
> the memory through insight 5.0 but when i connect to target and issue a
> download the whole environment hangs. Has it anything to do with my ecos
> configuration or is it something else?

Have you tried just setting a single memory address in RAM to see if that
works, eg. set *(int *)XXXXXX=0x87654321

where XXXXX is some valid RAM address. Also try using the command line to
eliminate the possibility of it being an insight-specific problem.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Jesper Skov]

>>>>> "Daniel" == Daniel Andersson <Daniel.Andersson@combitechsystems.com> writes:

Daniel> Hi, I have just recently got my EB40 board on-line with gdb

What is a EB40 board?

Daniel> (using serial/angel). However, i cannot download anything to
Daniel> the memory. I can view the memory through insight 5.0 but when
Daniel> i connect to target and issue a download the whole environment
Daniel> hangs. Has it anything to do with my ecos configuration or is
Daniel> it something else?

Uh, more likely that there's no eCos port for the platform. What
target did you compile for?

Jesper

RE: [ECOS] Problem downloading eCos

[Daniel.Andersson]

I just tried to set RAM using insight and it works fine. I can alter any
memory position that i want so unfortunately i dont think that is the
problem.

Maybe you have a very small file that i can try to download just to see if
its anything wrong with my download file. The current file that i am trying
to download is almost 1Mb and it seems a bit strange since its only a hello
world application. This file was built together with eCos ARM PID default
settings.

/Daniel

> -----Original Message-----
> From: Jonathan Larmour [ mailto:jlarmour@redhat.com ]
> Sent: den 20 mars 2001 16:53
> To: Andersson Daniel
> Cc: ecos-discuss@sourceware.cygnus.com
> Subject: Re: [ECOS] Problem downloading eCos
> 
> 
> Daniel.Andersson@combitechsystems.com wrote:
> > 
> > Hi,
> > 
> > I have just recently got my EB40 board on-line with gdb (using
> > serial/angel). However, i cannot download anything to the 
> memory. I can view
> > the memory through insight 5.0 but when i connect to target 
> and issue a
> > download the whole environment hangs. Has it anything to do 
> with my ecos
> > configuration or is it something else?
> 
> Have you tried just setting a single memory address in RAM to 
> see if that
> works, eg. set *(int *)XXXXXX=0x87654321
> 
> where XXXXX is some valid RAM address. Also try using the 
> command line to
> eliminate the possibility of it being an insight-specific problem.
> 
> Jifl
> -- 
> Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 
> (1223) 271062
> Maybe this world is another planet's Hell -Aldous Huxley || 
> Opinions==mine
> 

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

Daniel.Andersson@combitechsystems.com wrote:
> 
> I just tried to set RAM using insight and it works fine. I can alter any
> memory position that i want so unfortunately i dont think that is the
> problem.
> 
> Maybe you have a very small file that i can try to download just to see if
> its anything wrong with my download file. The current file that i am trying
> to download is almost 1Mb and it seems a bit strange since its only a hello
> world application. This file was built together with eCos ARM PID default
> settings.

Um, ARM PID? But you said your board is an EB40!

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Jonathan,

> > world application. This file was built together with eCos ARM PID default
>
>Um, ARM PID? But you said your board is an EB40!

To the best of your knowledge, did anyone officially pick up the 
unsupported AT91EB01 port that was mentioned in the list some time ago?


=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> Hi Jonathan,
> 
> > > world application. This file was built together with eCos ARM PID default
> >
> >Um, ARM PID? But you said your board is an EB40!
> 
> To the best of your knowledge, did anyone officially pick up the
> unsupported AT91EB01 port that was mentioned in the list some time ago?

I've kept the link in my pile of "net things to deal with", but I haven't
actively done anything yet - Carl did say it was quite old so it's not
something I expect to just be able to pick up and use.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

RE: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

>its anything wrong with my download file. The current file that i am trying
>to download is almost 1Mb and it seems a bit strange since its only a hello
>world application. This file was built together with eCos ARM PID default

When you say it's 1Mb, do you mean that the .exe is 1Mb long, or the actual 
code? Use arm-elf-size to check how big the actual code is... the 
executable will have loads of debug information in it that isn't uploaded 
to the board.

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

RE: [ECOS] Problem downloading eCos

[Daniel.Andersson]

Hi again,

Yes, the board is an AT91 evaluation board from ATMEL with an EB40 mounted
on it. The precessor is an AT91R40807. However, i thought that the option
ARM PID in the eCos configuration tool covered several of the ARM boards. Do
you mean that there are no port of eCos for my platform? And if eCos can run
on my platform how do i set up eCos to suite this board (if not to use the
PID option)?

/Daniel

> -----Original Message-----
> From: Jonathan Larmour [ mailto:jlarmour@redhat.com ]
> Sent: den 20 mars 2001 17:12
> To: Andersson Daniel
> Cc: ecos-discuss@sourceware.cygnus.com
> Subject: Re: [ECOS] Problem downloading eCos
> 
> 
> Daniel.Andersson@combitechsystems.com wrote:
> > 
> > I just tried to set RAM using insight and it works fine. I 
> can alter any
> > memory position that i want so unfortunately i dont think 
> that is the
> > problem.
> > 
> > Maybe you have a very small file that i can try to download 
> just to see if
> > its anything wrong with my download file. The current file 
> that i am trying
> > to download is almost 1Mb and it seems a bit strange since 
> its only a hello
> > world application. This file was built together with eCos 
> ARM PID default
> > settings.
> 
> Um, ARM PID? But you said your board is an EB40!
> 
> Jifl
> -- 
> Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 
> (1223) 271062
> Maybe this world is another planet's Hell -Aldous Huxley || 
> Opinions==mine
> 

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

Daniel.Andersson@combitechsystems.com wrote:
> 
> Hi again,
> 
> Yes, the board is an AT91 evaluation board from ATMEL with an EB40 mounted
> on it. The precessor is an AT91R40807. However, i thought that the option
> ARM PID in the eCos configuration tool covered several of the ARM boards.

Alas no.

> Do
> you mean that there are no port of eCos for my platform? And if eCos can run
> on my platform how do i set up eCos to suite this board (if not to use the
> PID option)?

There is no port included in the standard eCos sources yet. I think you may
want to start here:
http://sources.redhat.com/ml/ecos-discuss/2001-01/msg00496.html

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[massey steven]

--- Jonathan Larmour <jlarmour@redhat.com> wrote:
> Daniel.Andersson@combitechsystems.com wrote:
> > 
> > Hi again,
> > 
> > Yes, the board is an AT91 evaluation board from ATMEL with an EB40 mounted
> > on it. The precessor is an AT91R40807. However, i thought that the option
> > ARM PID in the eCos configuration tool covered several of the ARM boards.
> 
> Alas no.
> 
> > Do
> > you mean that there are no port of eCos for my platform? And if eCos can run
> > on my platform how do i set up eCos to suite this board (if not to use the
> > PID option)?
> 
> There is no port included in the standard eCos sources yet. I think you may
> want to start here:
> http://sources.redhat.com/ml/ecos-discuss/2001-01/msg00496.html
> 
> Jifl
> -- 
> Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
> Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Hello All,

It's been a long while since i've worked with eCos.  But, i had it working fine on the
EB01/AT9140807 board, using Carl's & my updates.  Really it's pretty trivial.  But, I am
just now starting back on my project and can probably provide assistance.  Maybe, we can
finally get the EB01/EB40 platforms permanantly in the source tree??

Have Fun,
Steven Massey



__________________________________________________
Do You Yahoo!?
Get email at your own domain with Yahoo! Mail. 
http://personal.mail.yahoo.com/

RE: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Daniel,

>Yes, the board is an AT91 evaluation board from ATMEL with an EB40 mounted
>on it. The precessor is an AT91R40807. However, i thought that the option
>ARM PID in the eCos configuration tool covered several of the ARM boards. Do

The EB40 isn't supported right now. However, if you want to work on a port 
for this board, I would be most interested in pooling resources (as a 
personal project, not for my work).

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

RE: [ECOS] Problem downloading eCos

[Daniel.Andersson]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1472 bytes --]

Hi,

Since eCos isn´t supported then i don´t have any other options than porting
it because in my project we must use the AT91R40807 processor. And i would
be very greatful if i could use you as a panel in some porting issues. 

Is the EB01 eCos port stable at the moment? Can i use it and make the
appropriate changes to suite the EB40 or should i start from some other
source?

By the way, do you have a working "hello world"-example for the EB40 board
(without eCos). Just so a can get my board running.

/Daniel Andersson

> -----Original Message-----
> From: Lewin A.R.W. Edwards [ mailto:larwe@larwe.com ]
> Sent: den 21 mars 2001 14:54
> To: Andersson Daniel; jlarmour@redhat.com
> Cc: ecos-discuss@sourceware.cygnus.com
> Subject: RE: [ECOS] Problem downloading eCos
> 
> 
> Hi Daniel,
> 
> >Yes, the board is an AT91 evaluation board from ATMEL with 
> an EB40 mounted
> >on it. The precessor is an AT91R40807. However, i thought 
> that the option
> >ARM PID in the eCos configuration tool covered several of 
> the ARM boards. Do
> 
> The EB40 isn't supported right now. However, if you want to 
> work on a port 
> for this board, I would be most interested in pooling resources (as a 
> personal project, not for my work).
> 
> === Lewin A.R.W. Edwards (Embedded Engineer)
> Work: http://www.digi-frame.com/
> Personal: http://www.zws.com/ and http://www.larwe.com/
> 
> "Und setzet ihr nicht das Leben ein,
> Nie wird euch das Leben gewonnen sein."
> 

RE: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2272 bytes --]

Hi Daniel,

In response to this (and your other email)...

>Since eCos isn´t supported then i don´t have any other options than porting
>it because in my project we must use the AT91R40807 processor. And i would
>be very greatful if i could use you as a panel in some porting issues.

I'm working with the EB40 for a semi-personal project. Meaning, it is 
something I can't get official funding for, so I bought the eval kit and am 
working on it at home. Once I get the idea demonstrable, I will bring it to 
work and pitch it to management. It's much easier to push an idea to 
production if you can say "but all the work is already done..." :)

>Is the EB01 eCos port stable at the moment? Can i use it and make the
>appropriate changes to suite the EB40 or should i start from some other
>source?

I haven't used it. At the moment I am working with standalone code. I have 
taken the ARM PID HAL from the current CVS sources and I have started to 
work on a from-scratch port, but I don't know when it can be finished.

>By the way, do you have a working "hello world"-example for the EB40 board
>(without eCos). Just so a can get my board running.

OK, I will email you a simple LED blinking app in C (makefile, sourcefiles, 
linker script) privately. Please note that this code does _not_ do full C 
run-time startup suitable for ROMmed applications. Tauno Voipio (from 
comp.arch.embedded) sent me some much more detailed code for standalone 
ROMmed applications, but before I can use it I need to write a flashloader 
for the EB40 (Atmel doesn't supply a built tool for it).

Tauno is working with somewhat old tools configured rather differently from 
the tools used to build eCos. So I have had to do some adapting.

There are also some performance issues with eCos. Due to current gcc 
bug(s), it's not possible to compile eCos in Thumb mode. However, Thumb is 
_STRONGLY_ recommended on the AT91R40807 because of its 16-bit data path. I 
want to solve this issue (read: I want to work out how to un-break eCos 
when compiled in thumb mode).

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> There are also some performance issues with eCos. Due to current gcc
> bug(s), it's not possible to compile eCos in Thumb mode. However, Thumb is
> _STRONGLY_ recommended on the AT91R40807 because of its 16-bit data path. I
> want to solve this issue (read: I want to work out how to un-break eCos
> when compiled in thumb mode).

Assuming this is the same problem I found (compilation fails with "byte or
halfword not valid for base register"), then work on this:

http://sources.redhat.com/ml/binutils/2001-02/threads.html#00059

If you can produce a small enough testcase for Richard he may be able to
help. This got pushed down my stack and I never got to it (I've been saying
that a lot recently I know).

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Jonathan,

> > There are also some performance issues with eCos. Due to current gcc
> > bug(s), it's not possible to compile eCos in Thumb mode. However, Thumb is
>
>Assuming this is the same problem I found (compilation fails with "byte or
>halfword not valid for base register"), then work on this:

Yep, same problem.

> http://sources.redhat.com/ml/binutils/2001-02/threads.html#00059

That link gets me to a very long list of threads, none of which seems 
immediately relevant, but I just took a cursory look over the page.

>If you can produce a small enough testcase for Richard he may be able to

I'll _try_. At the moment I am still recovering from a reformat, so I'm 
still re-downloading CVS trees of half the software in the universe. "The 
statistical probability is that other civilizations will arise. There will 
some day be lemon-scented paper napkins. Until then, there will be a short 
delay. Please return to your seat".

>help. This got pushed down my stack and I never got to it (I've been saying
>that a lot recently I know).

It's OK, the bottom of my stack will never see the light of day. Could be a 
couple of prehistoric Cro-Magnon plates down there for all I know.

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> Hi Jonathan,
> 
> > > There are also some performance issues with eCos. Due to current gcc
> > > bug(s), it's not possible to compile eCos in Thumb mode. However, Thumb is
> >
> >Assuming this is the same problem I found (compilation fails with "byte or
> >halfword not valid for base register"), then work on this:
> 
> Yep, same problem.
> 
> > http://sources.redhat.com/ml/binutils/2001-02/threads.html#00059
> 
> That link gets me to a very long list of threads, none of which seems
> immediately relevant, but I just took a cursory look over the page.

It should be centred on one thread titled "ARM thumb byte stores"

Thanks for volunteering to give it a go!

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

> > > http://sources.redhat.com/ml/binutils/2001-02/threads.html#00059
> >
> > That link gets me to a very long list of threads, none of which seems
>
>It should be centred on one thread titled "ARM thumb byte stores"

Bizarre - I clicked that link again, and this time it got me to the right 
message. *shrug* I won't even try to explain it :)

I'm just updating my gcc sources to the latest right now, then I will 
rebuild and make this problem my project for today.

BTW, I'm a tad confused as to which configurations of which tools I should 
be using to achieve this. If I'm using the latest gcc (which has merged 
arm-thumb and arm-elf), then should I configure binutils 2.10 for arm-thumb 
or arm-elf? Or should I use cvs binutils?

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Jonathan,

> > I'm just updating my gcc sources to the latest right now, then I will
>
>arm-elf is certainly what I used, and I invoked arm-elf-gcc with -mthumb. I

Silly, silly, silly, stupid, forgetful me. I forgot that current versions 
of gcc are broken in cygwin. 2.95.2 is the last version that seems to build 
properly. Here's the output from the current sources.

So it isn't possible to develop for thumb inside cygwin, unless you're 
writing in 100% assembler.

gcc -DCROSS_COMPILE -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings 
-Wstrict-prototy
pes -Wmissing-prototypes -Wtraditional -pedantic 
-Wno-long-long  -DHAVE_CONFIG_H
  -DGENERATOR_FILE  -o gencodes.exe \
  gencodes.o rtl.o bitmap.o ggc-none.o gensupport.o hashtab.o safe-ctype.o 
print-
rtl.o errors.o ` case "obstack.o" in ?*) echo obstack.o ;; esac ` ` case "" 
in ?
*) echo  ;; esac ` ` case "" in ?*) echo  ;; esac ` ` case "" in ?*) 
echo  ;; es
ac ` -ladvapi32
./gencodes.exe /src/gcc/gcc/config/arm/arm.md > tmp-codes.h
'src/gcc/gcc/config/arm/arm.md:36: unknown rtx code `define_constants
/src/gcc/gcc/config/arm/arm.md:36: following context is 
`  [(IP_REGNUM      12);
  Scratch register'
make[1]: *** [s-codes] Error 1
make[1]: Leaving directory `/tmp/build/gcc/gcc'
make: *** [all-gcc] Error 2
make: Leaving directory `/tmp/build/gcc'

administrator@LARWEP3 /tmp/build/gcc
$
=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> Hi Jonathan,
> 
> > > I'm just updating my gcc sources to the latest right now, then I will
> >
> >arm-elf is certainly what I used, and I invoked arm-elf-gcc with -mthumb. I
> 
> Silly, silly, silly, stupid, forgetful me. I forgot that current versions
> of gcc are broken in cygwin. 2.95.2 is the last version that seems to build
> properly. Here's the output from the current sources.
[snip]
> ./gencodes.exe /src/gcc/gcc/config/arm/arm.md > tmp-codes.h
> 'src/gcc/gcc/config/arm/arm.md:36: unknown rtx code `define_constants
> /src/gcc/gcc/config/arm/arm.md:36: following context is
> `  [(IP_REGNUM      12);
>   Scratch register'
> make[1]: *** [s-codes] Error 1
> make[1]: Leaving directory `/tmp/build/gcc/gcc'
> make: *** [all-gcc] Error 2
> make: Leaving directory `/tmp/build/gcc'

Looks suspiciously like our old friend the CRLF problem. It's barfed on the
very first non-comment line.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

>Looks suspiciously like our old friend the CRLF problem. It's barfed on the
>very first non-comment line.

I think I may have done something very stupid, please stand by for an hour 
while I download the sources again.

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

> > ./gencodes.exe /src/gcc/gcc/config/arm/arm.md > tmp-codes.h
>
>Looks suspiciously like our old friend the CRLF problem. It's barfed on the
>very first non-comment line.

Hmm, I find out that what I thought was a stupid mistake on my part isn't 
after all.

What gives? I load that file in vi and it tells me [dos]. Yet, it tells me 
the same thing for all the other files, and they don't cause problems. I 
use tr -d '\r' and now the compile process goes further (it's running while 
I type this).

Why does this one file cause problems? Does "gencodes.exe" open the machine 
descriptor file in binary mode?? If so, it's broken for cygwin. Is it 
something that's supplied with cygwin, or part of the gcc distribution?

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> > > ./gencodes.exe /src/gcc/gcc/config/arm/arm.md > tmp-codes.h
> >
> >Looks suspiciously like our old friend the CRLF problem. It's barfed on the
> >very first non-comment line.
> 
> Hmm, I find out that what I thought was a stupid mistake on my part isn't
> after all.
> 
> What gives? I load that file in vi and it tells me [dos]. Yet, it tells me
> the same thing for all the other files, and they don't cause problems.

Because CVS versions of gcc now ignore carriage returns. Old versions of
gcc wouldn't fare so well. But gencodes just reads the file (in normal text
mode).

> Why does this one file cause problems? Does "gencodes.exe" open the machine
> descriptor file in binary mode?? If so, it's broken for cygwin. Is it
> something that's supplied with cygwin, or part of the gcc distribution?

Part of gcc. No, this is down to CVS checking out the files with ^Ms. 

Firstly, make sure you're using cygwin CVS and not another one. And
secondly, here's a similar post from the cygwin list. Point 2a) is the
interesting one, although I've heard people report that having it mounted
in text mode doesn't work either. Or maybe it was only the build directory
in text mode that fails.

-=-=-=-=-

--- Christopher Cobb <ccobb@cseg.com> wrote:
> I am getting whole-file conflicts where the only difference is the
> presence of ^M's in one version and not in the other.
> 
> Does this have to do with how the end-of-file setting is set in cygwin?
> 
> We have unix/linux clients and windows clients (with cygwin).  Cygwin
> and non-cygwin tools are used to edit files (i.e., tools which put ^M's
> into the files).
> 
> How can I avoid whole-file conflicts?
> 

Three methods:

1) Filter the text files so that the ^M is removed.
   a) Rename cvs.exe to cygcvs.exe
   b) Create a script named cvs that filters the files in a manner
appropriate
      for you.  There are several ways to do it, none of which is more
right
      than the other.
   c) execute cygcvs with the options passed to your script.

2) Remount you Cygwin mounts in text mode.
   a) mount -f c:/cygwin /          # you didn't give the `cygcheck -r -s
-v'
                                    # output so I don't know what mounts
*you*
                                    # have now.
   b) If you have true binary files being checked in and out of CVS then
this
      won't work for you.

3) Get the cvs sources from the Cygwin mirror and rebuild it with
automode.o
   a) configure --prefix=/usr --sysconfdir=/etc --libexecdir=/sbin
   b) modifiy the Makefile so that the cvs.exe is created with the 
      /usr/lib/automode.o object
   c) make
   d) make install

Automode.o is a read in text, write in binary.  The alternative to this is
to
change the code in CVS.  If you're updating binary files this might not
work
for you.

The three options above may or may not work.  They are suggestions to try
not
definitive methods of cure.

Cheers,

=====
Earnie Boyd
-=-=-=-=-

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

> > Hmm, I find out that what I thought was a stupid mistake on my part isn't
>[...]
> > Why does this one file cause problems? Does "gencodes.exe" open the machine
> > descriptor file in binary mode?? If so, it's broken for cygwin. Is it
>Firstly, make sure you're using cygwin CVS and not another one. And

This was the dumb mistake I thought I'd made, but it turns out I didn't 
make that mistake. No, I was definitely, positively, absolutely using the 
cygwin CVS client. (So, why did it give me a File'O'^Ms? The behavior 
observed is consistent with CVS downloading without ^Ms, the text-mode 
mount point inserting them, and gencode.exe opening the file in binary mode 
and seeing them)

>secondly, here's a similar post from the cygwin list. Point 2a) is the
>interesting one, although I've heard people report that having it mounted
>in text mode doesn't work either. Or maybe it was only the build directory
>in text mode that fails.

As you can see from my second post, manually translating the arm.md file 
"fixed" that problem but leads me to a link problem.

The build directory is mounted in binary mode, the source directory is 
mounted in text mode.

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Jonathan,

>Looks suspiciously like our old friend the CRLF problem. It's barfed on the

Well, still no worky:



gcc -DCROSS_COMPILE -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings 
-Wstrict-prototy
pes -Wmissing-prototypes -Wtraditional -pedantic 
-Wno-long-long  -DHAVE_CONFIG_H
   -o cc1.exe \
         c-parse.o c-lang.o c-errors.o c-lex.o c-pragma.o c-decl.o 
c-typeck.o c-c
onvert.o c-aux-info.o c-common.o c-format.o c-semantics.o c-dump.o 
libcpp.a  mai
n.o libbackend.a obstack.o  -lintl     -ladvapi32 ../libiberty/libiberty.a
libbackend.a(recog.o)(.text+0x231b): undefined reference to `PUSH_ROUNDING'
libbackend.a(recog.o)(.text+0x2335): undefined reference to `PUSH_ROUNDING'
collect2: ld returned 1 exit status
make[1]: *** [cc1.exe] Error 1
make[1]: Leaving directory `/tmp/build/gcc/gcc'
make: *** [all-gcc] Error 2
make: Leaving directory `/tmp/build/gcc'


=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> Hi Jonathan,
> 
> >Looks suspiciously like our old friend the CRLF problem. It's barfed on the
> 
> Well, still no worky:
> 
> gcc -DCROSS_COMPILE -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings
> -Wstrict-prototy
> pes -Wmissing-prototypes -Wtraditional -pedantic
> -Wno-long-long  -DHAVE_CONFIG_H
>    -o cc1.exe \
>          c-parse.o c-lang.o c-errors.o c-lex.o c-pragma.o c-decl.o
> c-typeck.o c-c
> onvert.o c-aux-info.o c-common.o c-format.o c-semantics.o c-dump.o
> libcpp.a  mai
> n.o libbackend.a obstack.o  -lintl     -ladvapi32 ../libiberty/libiberty.a
> libbackend.a(recog.o)(.text+0x231b): undefined reference to `PUSH_ROUNDING'
> libbackend.a(recog.o)(.text+0x2335): undefined reference to `PUSH_ROUNDING'
> collect2: ld returned 1 exit status

Build failure of the moment probably. Time to read
http://gcc.gnu.org/bugs.html and submit it into GNATS.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

> > libbackend.a(recog.o)(.text+0x231b): undefined reference to `PUSH_ROUNDING'
> > libbackend.a(recog.o)(.text+0x2335): undefined reference to `PUSH_ROUNDING'
> > collect2: ld returned 1 exit status
>
>Build failure of the moment probably. Time to read
> http://gcc.gnu.org/bugs.html and submit it into GNATS.

OK, I've done that. In the meantime I will try to upgrade the native gcc on 
the system. Maybe the newer snapshots can't be built with the older gcc.


=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

> > libbackend.a(recog.o)(.text+0x2335): undefined reference to `PUSH_ROUNDING'
> > collect2: ld returned 1 exit status
>
>Build failure of the moment probably. Time to read

It turns out that this was a bug both introduced and fixed the day I 
downloaded my sources (typical luck, eh?). There are several problems 
lurking behind this one, though. However after some hours tweaking I have 
got a lot closer to getting it to compile. Maybe you can help? I can't get 
past these errors at the end. I think I'm getting near though.

Below is what to do. I am not sure that all of those steps are necessary. 
However they do point to some broken parts of the build process. In 
particular, I don't understand why those header files get included 
recursively in cygwin, but not in Linux...



Download gcc sources from 20010322 or later (NOTE: 20010321 has a known 
problem, don't download it). Configure gcc according to instructions for 
arm-elf. The text below assumes you have your gcc sources in 
/src/gcc/gcc-xxxxxxxx and your temp build directory is /tmp/build/gcc.

Strip CRs from /src/gcc/gcc-xxxxxxxx/gcc/config/arm/arm.md

Build according to instructions. You'll eventually get to about 10,000 
lines of errors. Strip CRs from /tmp/build/gcc/gcc/tconfig.h and edit that 
file:

add to the top:
#ifndef LARWE_TCONFIG_H
#define LARWE_TCONFIG_H

add to the bottom:
#endif

Strip CRs from the file /src/gcc/gcc-xxxxxxxx/gcc/config/arm/elf.h, edit 
the file and add to the top:
#ifndef LARWE_ARM_ELF_H
#define LARWE_ARM_ELF_H

add to the bottom:
#endif

Strip CRs from the file /src/gcc/gcc-xxxxxxxx/gcc/config/arm/unknown-elf.h, 
edit the file and add to the top:
#ifndef LARWE_UNKNOWN_ELF_H
#define LARWE_UNKNOWN_ELF_H

add to the bottom:
#endif

Strip CRs from the file /src/gcc/gcc-xxxxxxxx/gcc/longlong.h, edit the file 
and add a blank line to the end.


Now you wind up with these errors (see make.out attached)

Re: [ECOS] Problem downloading eCos

[Jonathan Larmour]

"Lewin A.R.W. Edwards" wrote:
> 
> 
> Now you wind up with these errors (see make.out attached)

This seems odd:

/digiframe/gccdl/gcc/gcc/longlong.h:2236:5: warning: no newline at end of
file

It has a newline in my sources. I think this may be causing it to lose the
#endif on that line. Add a newline and see what happens.

Some of the rest of the line numbers seem wrong so don't read too much into
them.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] Problem downloading eCos

[Lewin A.R.W. Edwards]

Hi Jonathan,

>/digiframe/gccdl/gcc/gcc/longlong.h:2236:5: warning: no newline at end of
>file

I already put two blank lines at the end of this file. Something's screwed...

=== Lewin A.R.W. Edwards (Embedded Engineer)
Work: http://www.digi-frame.com/
Personal: http://www.zws.com/ and http://www.larwe.com/

"Und setzet ihr nicht das Leben ein,
Nie wird euch das Leben gewonnen sein."