rtems-libbsd: Thoughts on Refactoring linkcmds.base

[Thomas Doerfler]

Sebastian,

will there also be architecture specific linkcomds, which will collect
sections defined by GCC? AFAIK, GCC defines different sections (and they
change with GCC evolution) for each architecture, so IMHO it would be
good to have separate linkcmds fragments for:
- BSP specifics
- GCC/architecture specifics
- RTEMS specifics

Any ideas on that?

wkr,

Thomas.


Am 25.07.2012 10:04, schrieb Sebastian Huber:
> On 07/25/2012 09:22 AM, Kevin Polulak wrote:
>> On Mon, Jul 23, 2012 at 3:08 AM, Sebastian Huber
>> <sebastian.huber at embedded-brains.de
>> <mailto:sebastian.huber at embedded-brains.de>> wrote:
>>
>>     A short term solution is to just create the linkcmds.rwrtems and
>>     linkcmds.rortems and use them in selected BSPs.
>>
>>
>> I'm confused. What are these files for? What's rwrtems and rortems?
> 
> These files should be used to define all input sections which are RTEMS
> specific.  One should contain all read-write sections.  One should
> contain all read-only sections.
> 
>>
>> On Mon, Jul 23, 2012 at 9:14 AM, Joel Sherrill <joel.sherrill at oarcorp.com
>> <mailto:joel.sherrill at oarcorp.com>> wrote:
>>
>>     + Some architectures get an underscore prepended. So
>>          symbols in the linkcmds get one more _ in front than
>>          you see in C.
>>     + The case of the sparc where there was an end text
>>          type symbol which really marked the end of where
>>          constant stuff like .text, rodata, .ctors tables, etc. was.
>>     + Where are the BSD sections located (e.g. >ram)
>>
>>     I think the first can be dealt with by either having two
>>     versions of the linkcmds.xxx or having both styles of
>>     symbols. We have had both in the past especially for
>>     BSPs which transitioned from a.out to coff to elf.
>>
>>
>> I wasn't aware that the prepending of underscores was
>> architecture-dependent. I
>> thought that was just a GNU ld thing and that it /always/ prepended one?
>>
>> Having two versions of linkcmds.xxx would be the most clean. That way the
>> symbol tables aren't being polluted with useless symbols. However, if
>> just
>> defining both symbol and _symbol in a single file is what's been used
>> in the
>> past, then that's probably the way to go. For me, consistency always
>> trumps all.
> 
> I don't think we need two linkcmds.xxx.  The underscore issue should be
> sorted out in the header files.
> 
>>
>>     The second is a bit harder for me to see a general solution.
>>
>>
>> I'm not sure why this would be an issue. Are you referring to the
>> _endtext and
>> friends symbols? If so, can't they just be moved? I don't see the
>> problem with
>> .rodata, or .ctors since I never had to touch them but for .text I
>> just moved
>> it to where it belonged: at the end of .text.
> 
> I don't see an issue with these symbols too.
> 


-- 
--------------------------------------------
embedded brains GmbH
Thomas Doerfler
Obere Lagerstrasse 30
D-82178 Puchheim
Germany
email: Thomas.Doerfler at embedded-brains.de
Phone: +49-89-18908079-2
Fax: +49-89-18908079-9
PGP: Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.