[Bug gas/15776] MSP430 "add" in "Rn, Rm" mode sometimes assembles to nothing

[quandary at remstate dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15776

Travis Snoozy <quandary at remstate dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
            Summary|MSP430 "add" and "addc" in  |MSP430 "add" in "Rn, Rm"
                   |"Rn, Rm" mode assembles to  |mode sometimes assembles to
                   |nothing                     |nothing

--- Comment #3 from Travis Snoozy <quandary at remstate dot com> ---
I sync'd and built HEAD yesterday, so I'm on the latest there is.

I found the discreptancy that made me think this was an easy repro: the first
line in the .S file does not show up properly in the listing file. I was going
by the listing file's output, not objdump, and my tests always started with the
"known bad" opcode. Putting in the ".text" leader worked around that problem,
and invalidated my test cases (read: the listing files come out as they should
now). So, that explains why my repro... well, doesn't repro properly. ;)


However, my actual (non-test) program *is* having all four of its (Rn, Rm) mode
adds disappear. In one of the four cases (the others are pretty much identical)
the listing files shows:

<...>
  76                .LtrackingLabel:
  77
  78
  79 0070 7946          mov.b   @R6+,       R9
  80                add R7,     R9
  81 0072 6949          mov.b   @R9,        R9
<...>

Note the gap. In the objdump output:

<...>
0000e4e6 <.LtrackingLabel>:
    e4e6:       79 46           mov.b   @r6+,   r9
    e4e8:       69 49           mov.b   @r9,    r9
<...>

No add, broken program. However, with *just* the code snippets above, I can't
repro the problem.

I'll poke around a bit more and update this if I can make a working toy repro.

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