Re: [microblaze-uclinux] uClibc vs. newlib

[John Williams <jwilliams@xxxxxxxxxxxxxx>]

Hi John,

John Carter wrote:

> 1. I assume that we need to build a toolchain with uClibc rather than 
> newlib.  Am I missing something?  It looks like uClibc (at least the 
> version under ucLinux-dist in CVS) can be configured to build a 
> microblaze version.

To clarify the two different approaches:

(1) build mb-gcc against newlib.  Current mb-gcc distributions are done 
this way.  The uClinux-dist build process handles linking against uClibc 
(which itself gets built as part of the uClinux-dist Makefile).

(2) build a dedicated uClinux-targetting mb-gcc, which defaults to link 
against uClibc.  Some of the m68k (and arm) toolchains for uClinux do 
this.

Option (1) is easier, and would require much less hacking on the 
already-difficult (impossible?) mb-gcc build process.  It also means 
that mb-gcc can still be used to build executables targetting 
"bare-metal" microblaze programs, as well as uClinux-based apps (e.g. 
the bootloader).

Option (2) is nice if you want to build microblaze-uclinux applications 
outside the uClinux-dist approach.  Some of the m68k- and arm-gcc 
toolchains distributed by SnapGear do this.

My personal opinion is that it's so easy to include user apps in a 
uClinux-dist build that I wouldn't bother with option 2 - it will be a 
fair amount of work for little practical gain. I'd concentrate on 
getting option (1) going first - even building mb-gcc against newlib 
would be an achievement at this stage!  Maybe after that's sorted out, 
option (2) might be worth investigating.

Note I'm not a toolchain hacker by any means, so I might be overstating 
the effort required for option (2).

Regards,

John

___________________________
microblaze-uclinux mailing list
microblaze-uclinux@itee.uq.edu.au
Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
Mailing List Archive : http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/