[Buildroot] Non-x86 support question

[Buildroot] Non-x86 support question

[Michael S. Zick]

Just wondering, does buildroot get tested as a cross-compiler
building native toolchain?

There is a similar bug posted, which I think refers to building
the foreign, native, compiler.
At least that is my own problem.  ;)

In my case, gcc's configure is failing when it tries to test for
the presence of mpfr (which is present, and built) - - -
(No problem with gpm) 

I have captured the configure command being run and the test file,
then ran them "by hand" with -v the best I can figure out at that point is:

The "canned" test routine is test compiled with the working
cross-compiler - and it can't find the mpf header file.

But all of the gcc command line options seem correct to my eye, and the
search paths that -v reports that gcc is using seem correct.

A true puzzle, not a show-stopper here, anytime by 2010.02 would be fine.
This is gcc-4.3.whatever targeting mipsel.

Mike

[Buildroot] Non-x86 support question

[Peter Korsgaard]

>>>>> "Michael" == Michael S Zick <minimod@morethan.org> writes:

 Michael> Just wondering, does buildroot get tested as a cross-compiler
 Michael> building native toolchain?

The BR2_PACKAGE_GCC_TARGET? No, that isn't commonly used. That doesn't
mean that it shouldn't work ofcourse.

 Michael> There is a similar bug posted, which I think refers to building
 Michael> the foreign, native, compiler.
 Michael> At least that is my own problem.  ;)

 Michael> In my case, gcc's configure is failing when it tries to test for
 Michael> the presence of mpfr (which is present, and built) - - -
 Michael> (No problem with gpm) 

Hmm, it worked for me last time I tried (some time this summer).

 Michael> I have captured the configure command being run and the test file,
 Michael> then ran them "by hand" with -v the best I can figure out at
 Michael> that point is:

Could you provide a bit more details? What goes wrong?

 Michael> A true puzzle, not a show-stopper here, anytime by 2010.02
 Michael> would be fine.  This is gcc-4.3.whatever targeting mipsel.

It would be good if you could add a bugreport to bugzilla so we don't
forget about it.

-- 
Bye, Peter Korsgaard

[Buildroot] Non-x86 support question

[Michael S. Zick]

On Thu November 26 2009, Peter Korsgaard wrote:
> >>>>> "Michael" == Michael S Zick <minimod@morethan.org> writes:
> 
>  Michael> Just wondering, does buildroot get tested as a cross-compiler
>  Michael> building native toolchain?
> 
> The BR2_PACKAGE_GCC_TARGET? No, that isn't commonly used. That doesn't
> mean that it shouldn't work ofcourse.
> 
>  Michael> There is a similar bug posted, which I think refers to building
>  Michael> the foreign, native, compiler.
>  Michael> At least that is my own problem.  ;)
> 
>  Michael> In my case, gcc's configure is failing when it tries to test for
>  Michael> the presence of mpfr (which is present, and built) - - -
>  Michael> (No problem with gpm) 
> 
> Hmm, it worked for me last time I tried (some time this summer).
> 
>  Michael> I have captured the configure command being run and the test file,
>  Michael> then ran them "by hand" with -v the best I can figure out at
>  Michael> that point is:
> 
> Could you provide a bit more details? What goes wrong?
> 

One more note while waiting for me to document this in a true bug-report.

I considered handling gpm and mpf as "patches" to gcc - -
Since gcc will configure and build them if it finds them in its source tree.

I made up a gcc tarball along those lines and tried it - -
Same result - cross-compiler builds, native-compiler fails to configure
while testing for mpf header files (two or three missing and/or renamed symbols).

As a reminder: this is gcc-4.3.x targeting MIPS (32-R2, 24Kc, hard-float).
gcc-4.4.x is hopeless for MIPS at the moment, implements an ABI change for MIPS,
mostly in how structures are packed.

Mike