[ECOS] Ignore earlier dumb question, but here is another

[ECOS] Ignore earlier dumb question, but here is another

[Lewin A.R.W. Edwards]

I think I'm on crack. It took about three hours for my messages to get back 
to me via the list and when they did, I immediately had a memory jolt that 
told me the correct switch to use. I've recompiled eCos with -fpack-struct 
and all is seemingly well. However I worry that the OS might not have been 
tested in this configuration. Any known caveats?
=== 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] Ignore earlier dumb question, but here is another

[Gary Thomas]

On 08-Jan-2001 Lewin A.R.W. Edwards wrote:
> I think I'm on crack. It took about three hours for my messages to get back 
> to me via the list and when they did, I immediately had a memory jolt that 
> told me the correct switch to use. I've recompiled eCos with -fpack-struct 
> and all is seemingly well. However I worry that the OS might not have been 
> tested in this configuration. Any known caveats?

I have tried it when building network code.  It doesn't break antyhing, AFAIK,
but it also may not always do just what you want.  YMMV

RE: [ECOS] Ignore earlier dumb question, but here is another

[Gary Thomas]

On 08-Jan-2001 Lewin A.R.W. Edwards wrote:
> I think I'm on crack. It took about three hours for my messages to get back 
> to me via the list and when they did, I immediately had a memory jolt that 

BTW this was a networking problem in California (USA), presumably brought on
by a large rain storm on Sunday-Monday.

RE: [ECOS] Ignore earlier dumb question, but here is another

[Lewin A.R.W. Edwards]

Hi Gary,

> > told me the correct switch to use. I've recompiled eCos with -fpack-struct
> > and all is seemingly well. However I worry that the OS might not have been
> > tested in this configuration. Any known caveats?
>
>I have tried it when building network code.  It doesn't break antyhing, AFAIK,
>but it also may not always do just what you want.  YMMV

Do you by chance have an alternate suggestion? Really the only part of my 
code that needs this functionality is the filesystem (and my JPEG codec). 
There must be a mechanism for accomplishing what I need.

I did try a "composite build" wherein my code was compiled with packed 
structures and eCos was compiled with aligned structures, and it DID work, 
but that's probably because I'm not using any OS functionality at all.

=== 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] Ignore earlier dumb question, but here is another

[Jesper Skov]

>>>>> "Lewin" == Lewin A R W Edwards <larwe@larwe.com> writes:

Lewin> I think I'm on crack. It took about three hours for my messages
Lewin> to get back to me via the list and when they did, I immediately
Lewin> had a memory jolt that told me the correct switch to use. I've
Lewin> recompiled eCos with -fpack-struct and all is seemingly
Lewin> well. However I worry that the OS might not have been tested in
Lewin> this configuration. Any known caveats?

Why don't you use the packed attribute on the structures that must be
packed, and allow the compiler to use aligment for better performance
on the rest?  See the GCC manual for information about the packed
attribute.

Jesper

RE: [ECOS] Ignore earlier dumb question, but here is another

[Gary Thomas]

On 09-Jan-2001 Lewin A.R.W. Edwards wrote:
> Hi Gary,
> 
>> > told me the correct switch to use. I've recompiled eCos with -fpack-struct
>> > and all is seemingly well. However I worry that the OS might not have been
>> > tested in this configuration. Any known caveats?
>>
>>I have tried it when building network code.  It doesn't break antyhing, AFAIK,
>>but it also may not always do just what you want.  YMMV
> 
> Do you by chance have an alternate suggestion? Really the only part of my 
> code that needs this functionality is the filesystem (and my JPEG codec). 
> There must be a mechanism for accomplishing what I need.
> 
> I did try a "composite build" wherein my code was compiled with packed 
> structures and eCos was compiled with aligned structures, and it DID work, 
> but that's probably because I'm not using any OS functionality at all.

Look at the networking code.  We use __attribute__(pack) on the structures
which matter and it works just fine.  It avoids the "big hammer" which often
does not function as intended.