[U-Boot] [PATCH 1/2 V2] Move libgcc inclusion from common Makefile to platform configs files

[Mike Frysinger]

On Saturday 20 June 2009 09:57:51 Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 09:33 Sat 20 Jun     , Mike Frysinger wrote:
> > On Saturday 20 June 2009 09:01:36 Jean-Christophe PLAGNIOL-VILLARD wrote:
> > > On 08:57 Sat 20 Jun     , Mike Frysinger wrote:
> > > > On Saturday 20 June 2009 07:30:41 Jean-Christophe wrote:
> > > > > On 07:08 Sat 20 Jun     , Mike Frysinger wrote:
> > > > > > On Saturday 20 June 2009 06:40:07 Jean-Christophe wrote:
> > > > > > > On 06:18 Sat 20 Jun     , Mike Frysinger wrote:
> > > > > > > > On Saturday 20 June 2009 05:33:26 Jean-Christophe wrote:
> > > > > > > > > This patch moves the libgcc Makefile inclusion from the
> > > > > > > > > toplevel Makefile to the arch_config.mk files. This is in
> > > > > > > > > preparation for the ARM architecture to move away from
> > > > > > > > > including libgcc function and only using self-contained
> > > > > > > > > U-Boot functions as done in Linux.
> > > > > > > >
> > > > > > > > why not change the top level Makefile to read:
> > > > > > > > PLATFORM_LIBGCC ?= ...
> > > > > > > >
> > > > > > > > then any board/arch that doesnt want it can simply do:
> > > > > > > > PLATFORM_LIBGCC = # dont want it
> > > > > > >
> > > > > > > because you need to provide the equivalent functions for
> > > > > > > standalone application and api and U-Boot ofcourse
> > > > > >
> > > > > > so move it to config.mk.  this doesnt change the important point:
> > > > > > leave PLATFORM_LIBGCC default in the toplevel common files.  what
> > > > > > i proposed doesnt limit what you want to do with arm in any way.
> > > > >
> > > > > I think it's better to let it at arch level and force any new arch
> > > > > adding to manage it instead provide a default one
> > > >
> > > > considering most arches want this code, i dont think so.  you're
> > > > proposing we duplicate the same common code in all arches to support
> > > > one deviating arch -- arm.  the defaults should reflect the common
> > > > state while the deviating ones change the behavior as they like.
> > >
> > > This is where I disagree other arch as mips and other will also need to
> > > move away from the needs to include libgcc to be really toolchain
> > > independant
> > >
> > > I do think it's really important for U-Boot be able to have full
> > > control to have a functions embedded into.
> > >
> > > So stop to have libgcc include by default will really reflect it
> >
> > so you have two arches (mips/arm) that you dont want to use libgcc.  that
> > is still vastly the minority.  if we ever do get most ports not using
> > libgcc, then pushing it to the arch configs makes sense.  but we havent
> > and we arent even close.
>
> which minoroty?

the majority use libgcc

> I just give you example but you will see arm, mips, sh, powerpc etc...
> toolchains problem just because we use the libgcc

4 is still the minority, but this is Woflgang's call

> I've patch in qualification for mips and sh so it will not be so long
> maybe 2 or 3 weeks

considering my proposed change is much simpler, and would allow for trivial 
migration as you converted things, it still makes more sense.

> > how you want to manage libgcc dependency is your prerogative ... it
> > doesnt mean you should be forcing other people to follow suit.
>
> you can do your own choice but I steel think it's a good improvement

it depends on the arch.  for ones doing stupid things (like arm), thats one 
thing.  but for others, it's simply a waste of time and unnecessary overhead.  
i'm not about to *copy* the exact sources from gcc just to avoid -lgcc.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20090620/52ae8c6e/attachment.pgp