Re: missing __init_array_start, __init_array_end in __libc_csu_init

[Michael Schmitz <schmitz@mail.biophys.uni-duesseldorf.de>]

> > hello.c: In function 'main':
> > hello.c:3: warning: incompatible implicit declaration of built-in function 'printf'
> > hello.c:2: warning: return type of 'main' is not 'int'
> > /usr/lib/libc_nonshared.a(elf-init.oS): In function `__libc_csu_init':
> > (.text+0x2e): undefined reference to `__init_array_start'
> > /usr/lib/libc_nonshared.a(elf-init.oS): In function `__libc_csu_init':
> > (.text+0x36): undefined reference to `__init_array_end'
> > /usr/bin/ld: a.out: hidden symbol `__init_array_end' isn't defined
> > /usr/bin/ld: final link failed: Nonrepresentable section on output
> > collect2: ld returned 1 exit status
>
> I can confirm that this is a regression from etch.

[...]

> It also occurs in current sid with binutils_2.17cvs20070426-3. -1 failed
> to build, -2 lived 3-1/2 days, so we probably don't have a copy of it
> anywhere.

http://snapshot.debian.net/archive/2007/04/28/debian/pool/main/b/binutils/
has it. Same result.

> You can't install etch bintuils on sid or sid binutils on etch due to
> libc6 so I couldn't do any further regression testing.
>
> Anything else I can try?

On hobbes I get:

root@hobbes:/# nm usr/lib/libc_nonshared.a

elf-init.oS:
         U _GLOBAL_OFFSET_TABLE_
         U __init_array_end
         U __init_array_start
00000000 T __libc_csu_fini
00000008 T __libc_csu_init
         U _init

which is what I also get on powerpc. I doubt the HPPA bug was related
(there, the __libc_csu_fini and __libc_csu_init pointers were broken.
Here, binutils seems to fail to emit __init_array_end and
__init_array_start instead)

We can try to build a pre-2.17cvs version on sid, right? I'll try that on
washi ...

	Michael