[Bug binutils/31571] strip mangles 64-bit mach-o binaries

[nickc at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=31571

Nick Clifton <nickc at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |nickc at redhat dot com

--- Comment #1 from Nick Clifton <nickc at redhat dot com> ---
(In reply to Jubilee Young from comment #0)
Hi Jubilee,

> This issue may be constrained to aarch64 macOS. 

This is going to be a very hard problem (for me) to diagnose.  Mainly because I
do not have access to an AArch64 MacOS system.  Are you willing to run some
tests to help track down the cause ?


> Part of the issue appears to be mangling the header. The green diff on the
> right is what happens after a binary is mangled by strip.

Generally speaking strip should not be mangling the header in this way,
especially not the magic number.  (I am not a MacOS expert, but I assume that
the magic number is important and should not be changed).

Judging by the change in magic number however it appears that strip has changed
the output from a 64-bit format (BFD_MACH_O_MH_MAGIC_64) to 32-bit format
(BFD_MACH_O_MH_MAGIC).  Which is obviously wrong.

Hmm, looking in the code, I think that I might see the problem.  It appears
that the BFD library does not support generating version 2 (0xFEEDFACF) format
output.  Take a look at the bfd_mach_o_arm64_mkobject() function in
bfd/mach-o-aarch64.c (around line 57).  I have no idea why this is the case. 
Nor how easy it would be to generate the version 2 format output.

Do you fancy doing some binutils hacking of your own and seeing if you can
either a) add support for the version 2 format or b) have version 2 format
input rejected so that strip will not silently corrupt binaries ?

Cheers
  Nick


Assuming that this is correct

-- 
You are receiving this mail because:
You are on the CC list for the bug.