i386 differ fails when diffing against code with old C++ header contents in the context?

[ChrisNonyminus]

See this scratch: https://decomp.me/scratch/RbmNx
(Note: the header in question is a preprocessed context of bstring.h from the Hewlett Packard C++ STL.)

Diff error: Error running asm-differ: failed to find address immediate for line 'add %al,(%eax)'

[simonlindholm]

I'm pretty sure this is the same issue as described here: https://discord.com/channels/897066363951128586/897075447534354462/1205257191481675836. To quote:

I took a look at the .o you sent me. It looks to me like objdump is buggy here. The file is a COFF, with (objdump -h, objdump -t):

  1 .text         00000100  00000000  00000000  000001cc  2**4
                  CONTENTS, ALLOC, LOAD, RELOC, READONLY, CODE
...
  8 .text         000000b0  00000000  00000000  000003fa  2**4
                  CONTENTS, ALLOC, LOAD, RELOC, READONLY, CODE

[  8](sec  2)(fl 0x00)(ty    0)(scl   6) (nx 0) 0x000000f2 $L42316
...
[ 39](sec  9)(fl 0x00)(ty    0)(scl   6) (nx 0) 0x000000a9 $L42321

(0 vs 1-indexing there it seems, but otherwise matches up). objdump -d then emits two "Disassembly of section .text" and includes $L42321 in both of them
I wonder if a COFF -> ELF conversion tool in front of objdump could help?
or if one of the .text sections could be renamed? hex editing code.o and replacing the first and third .text with .texa makes it dump correctly

Oddly enough though I cannot reproduce with .o produced by decomp.me's export button and my local objdump... I wonder why.

[ChrisNonyminus]

Another issue involving this:
This function matches, but asm-differ gives the wrong output. Note the seemingly "added" lines. https://decomp.me/scratch/wwqHU