why do i have an apple local C4 constructor in my arm-linux shared object symtable?

why do i have an apple local C4 constructor in my arm-linux shared object symtable?

[Ray Lehtiniemi]

hi all

i have found a C++ symbol in a shared object which does not want to
demangle using objdump, and upon investigation, i have some concerns:

i am working with a binary shared object from a vendor.  it was
apparently compiled using

 GCC: (GNU) 3.3.1 (MontaVista 3.3.1-3.0.10.0300532 2003-12-23)

for an arm architecture

  file format elf32-bigarm
  architecture: arm, flags 0x00000150:

the symbol in question is _ZN11sectionlistC4Ev.  using my shiny
new printout of the g++ ABI, i believe this should demangle to something
along the lines of 'sectionlist::sectionlist(void);', assuming
that the 'C4' is the 'APPLE LOCAL decloning' constructor i found
a patch for at

  http://www.opensource.apple.com/darwinsource/10.3/gcc-1495/gcc/cp/mangle.c

now, my question is:  since the symbol came from a shared object
cross-compiled from an x86 host for an arm xscale target, both
running linux, why on earth would i have an apple-specific cloning
constructor in the shared object?  should i be worried about this?
the armv5b-softfloat-linux-objdump i created using crosstool doesn't
want to demangle it, which is worrisome.

can someone please let me know if i can expect any trouble from
this apple-specific symbol, or how it might have come to be placed
into an arm shared object?

please cc: me directly as i'm not subscribed.


thanks!

-- 
----------------------------------------------------------------------
     Ray L   <rayl@mail.com>

Re: why do i have an apple local C4 constructor in my arm-linux shared object symtable?

[Ian Lance Taylor]

Ray Lehtiniemi <rayl@mail.com> writes:

> i am working with a binary shared object from a vendor.  it was
> apparently compiled using
> 
>  GCC: (GNU) 3.3.1 (MontaVista 3.3.1-3.0.10.0300532 2003-12-23)

...

> now, my question is:  since the symbol came from a shared object
> cross-compiled from an x86 host for an arm xscale target, both
> running linux, why on earth would i have an apple-specific cloning
> constructor in the shared object?  should i be worried about this?
> the armv5b-softfloat-linux-objdump i created using crosstool doesn't
> want to demangle it, which is worrisome.

You are getting that type of mangled name--constructor type 4--because
MontaVista gcc includes a patch to generate those mangled names.  That
patch has never been in FSF gcc.  If you check the MontaVista source
RPM, you'll see it in a patch which they apply to the FSF sources.

> can someone please let me know if i can expect any trouble from
> this apple-specific symbol, or how it might have come to be placed
> into an arm shared object?

I don't know for sure, but I wouldn't expect any problems.  If you do
have problems, talk to MontaVista.

Ian