RFH: MIPS ld breaking .eh_frame data ...

[David Daney <ddaney at avtrex dot com>]

I am seeing a wierd problem building libgcj with a mipsel-linux cross 
compiler.

mipsel-linux-ld --version
GNU ld version 060915 20060915

But I get the same results with 2.17 also.

mipsel-linux-gcj --version
mipsel-linux-gcj (GCC) 4.2.0 20060914 (experimental)

ld appears to be changing the augmentation data for some CIEs

For example, here is the data for _Jv_Throw which is part of libgcj.  If 
I look at the .eh_frame before linking

 mipsel-linux-readelf --debug-dump=frames exception.o
.
.
.
The section .eh_frame contains:

00000000 00000014 00000000 CIE
  Version:               1
  Augmentation:          "zP"
  Code alignment factor: 1
  Data alignment factor: -4
  Return address column: 31
  Augmentation data:     00 00 00 00 00

  DW_CFA_def_cfa: r29 ofs 0

.
.
.
0000009c 0000001c 000000a0 FDE cie=00000000 pc=00000328..000003cc
  DW_CFA_set_loc: 00000338
  DW_CFA_def_cfa_offset: 40
  DW_CFA_advance_loc: 12 to 00000344
  DW_CFA_offset: r16 at cfa-16
  DW_CFA_offset: r17 at cfa-12
  DW_CFA_offset: r31 at cfa-8
  DW_CFA_nop

As far as I can tell, this is all correct.

Now if I look at libgcj.so.8.0.0 after linking, I get:
.
.
.
00004084 00000018 00000000 CIE
  Version:               1
  Augmentation:          "zRP"
  Code alignment factor: 1
  Data alignment factor: -4
  Return address column: 31
  Augmentation data:     10 00 00 00 00 00

  DW_CFA_def_cfa: r29 ofs 0
  DW_CFA_nop
  DW_CFA_nop
.
.
.
00004124 0000001c 000000a4 FDE cie=00004084 pc=00c51888..00c5192c
  DW_CFA_set_loc: 026679ce
  DW_CFA_def_cfa_offset: 40
  DW_CFA_advance_loc: 12 to 026679da
  DW_CFA_offset: r16 at cfa-16
  DW_CFA_offset: r17 at cfa-12
  DW_CFA_offset: r31 at cfa-8
  DW_CFA_nop

The FDE data is unchanged, however its interpretation has changed due to 
 the added PC relative augmentation data in the CIE.  Note that the 
DW_CFA_set_loc in the first instruction pushes the PC out of the range 
of the function so that the unwinder completly ignores the rest of the 
FDE program causing a crash at runtime.

I used to be able to build (and run) libgcj with slightly older versions 
of binutils, and gcc.

Does this seem like a problem with ld?  Or is gcc creating something 
inconsistent here?

Thanks in advance,
David Daney.