Don't define USE_PT_GNU_EH_FRAME when using uClibc

[David Daney]

Richard Sandiford wrote:
> Jie Zhang <jzhang918@gmail.com> writes:
> 
>>David Daney wrote:
>>
>>>Jie Zhang wrote:
>>>
>>>>When building a cross compiler with --with-sysroot option,
>>>>inhibit_libc is not defined. GCC checks if PT_GNU_EH_FRAME can be used
>>>
>>>I think a better option is to defind inhibit_libc for uClibc.
>>>
>>
>>When I searched the mailing list, I found this
>><http://gcc.gnu.org/ml/gcc-patches/2003-05/msg02068.html>.
>>If we define inhibit_libc for uClibc as we did for newlib before, 
>>-fprofile-arcs / -fbranch-probabilities maybe cannot be used with uClibc.
> 
> 
> FWIW, we've been using a similar patch to Jie's on csl/coldfire-4_1,
> except we made it conditional on the target format being flat, not on
> the C library being uClibc.
> 
> Jie, is your patch also targetted at uClinux, or is this something
> that is needed for both ELF and flat?  In other words, does uClibc's
> (ELF) dynamic loader really not support PT_GNU_EH_FRAME?  If it doesn't
> now, I imagine it might in future.
> 
> Like you, I think that defining inhibit_libc is just working around the
> problem.  I don't see why it's right to define it for uClibc and not for
> glibc.

I will defer to others on this.

My patches were mainly to get DWARF exception handling working properly 
on mipsel-linux-uclibc.  I have not fully cleaned them up and it has 
also been a month or so since I looked at it, so it is entirely possible 
that my approach is not optimal or correct for bfin/uClinux.

David Daney.