Re: Error building x86 toolchain

[Davide Perlatti <ketama79 at gmail dot com>]

Thank you very much for the detailed explanation. I' ve build an i686
toolchain on 64 bit host machine with same configuration.
Thanks
Davide

2012/7/2 Yann E. MORIN <yann.morin.1998@free.fr>:
> Davide, Mirko, All,
>
> On Monday 02 July 2012 18:32:40 Davide Perlatti wrote:
>> Ok, sorry... I want to build the same toolchain i686 on 2 different
>> pc: 32 and 64 bit. I don't know if the parameters Emit assembly for
>> CPU and Tune for CPU must be the same on both machine.
>
> These options are for the _target_ machine (the machine the toolchain
> generates code for), not the _host_ machine (the machine on which the
> toolchain runs).
>
> So, whether you are on a 32-bit or a 64-bit host does not impact these
> settings.
>
> The settings Mirko suggested (or higher values) are the correct settings
> because NPTL requires at least a i486 CPU, and the default CPU for x86
> is i386 which lacks the required instrcutions to properly implement NPTL.
>
>> The obtained x-tool on 32 bit machine are ok, now i want to build the
>> same toolchain on a 64 bit machine. Always and only 32 bit toolchain.
>
> If you want the "same" toolchain to run on both machines, there re two
> options:
>
> 1- build a static toolchain on the 32-bit machine, and copy that toolchain
>    to the 64-bit machine.
>    This way, you ensure that it is the exact same toolchain that runs on
>    both machines, and thus that you'll get the exact same behaviour.
>    In this case, do not forget to install the -devel packages of your distro
>    (eg libc6-dev, libncurses5-dev and so on (on Debian, adapt to your own
>    distro package names).
>
> 2- build the toolchain on any of the two machines, and copy the .config file
>    to the other machine and re-build the toolchain there.
>    This way, you get a toolchain that is somewhat /optimised/ for each host,
>    and will generate code for the same target.
>    Except for highly convoluted issues, the toolchains should produce
>    identical ouput (although there are cases where the bitness of the host
>    _may_ have some impacts on the generated binaries, where the sections
>    and objects could be linked in different orders; I saw that only once on
>    a specific 64-bit host which would make the linker behave differently
>    than on a 32-bit host). But you should almost safe going this way.
>
> If identicity is a very high requirement of yours, then go for the static
> toolchain. Otherwise, choose (toss a dime as a last resort to choose! ;-) )
>
> Regards,
> Yann E. MORIN.
>
> --
> .-----------------.--------------------.------------------.--------------------.
> |  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
> | +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
> | +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
> | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
> '------------------------------^-------^------------------^--------------------'

--
For unsubscribe information see http://sourceware.org/lists.html#faq