Busybox make, modify my /dev/null on host

Christian MICHON christian.michon at gmail.com
Thu Feb 14 07:57:51 UTC 2008


On Thu, Feb 14, 2008 at 2:09 AM, Mike Frysinger <vapier at gentoo.org> wrote:
> On Wednesday 13 February 2008, Christian MICHON wrote:
>  > seriously, back to the issue. I understand from the thread:
>
>  you dont then
>
>
>  > 1) gcc used to work
>
>  the toolchain was always broken.  newer flags were added to the trylink script
>  which the older toolchain did not support.  the fact that newer checks
>  triggered the bug is irrelevant.  the toolchain was always broken.
>

it's not newer checks. it's at the linking stage it's failing.

a toolchain bootstrapped without libintl/gettext works fine! try it!

>
>  > It just happened to me too. But I'm sure my toolchain is not broken,
>  > because it's working fine usually.
>
>  wrong.  daily observations make no absolute statements.  just because you
>  havent managed to trigger the bug doesnt mean the bug magically does not
>  exist.
>

it's not a daily observation. I provided a real fix, that does not
involve a complete recompilation of the toolchain and solves the
problem seen in both bbox 1.8.2 and 1.9.0.

I have not tried 1.9.1 and your fix, since I believe the toolchain can
be adjusted if the end user needs libintl/gettext.

Beyond this libintl/gettext issue, I've quite a big regression list
that proves the toolchain is not broken... I use it with uclibc/glibc
and even on Sun solaris...

>
>  > my 2 cents. and this fix will be in the next release of DetaolB (based
>  > on old but reliable gcc-3.4.6, link in my signature)
>
>  ah but once again you'd be wrong.  gcc-3.4.6 clearly has the bug where it will
>  unlink /dev/null in the case of certain errors.  why dont you actually look
>  at the code.  the issue wasnt resolved in gcc proper until the gcc-4.1
>  series.  i know because i implemented the fixes in gcc-3.3, gcc-3.4, and
>  gcc-4.0 for Gentoo.
>  -mike
>

Do you have a patch then ? I'd be happy to see it. Or maybe you can
send me pointers to it ?

One last point: I was not talking on cross compiling. Native compiling only...
My bad if what I mentioned does not apply on cross compilers...

-- 
Christian
--
http://detaolb.sourceforge.net/, a linux distribution for Qemu with Git inside !



More information about the busybox mailing list