Re: MIPS16 GNU Assembler bug

[Shane Nay]

On Tuesday 16 January 2001 10:24, Nick Clifton wrote:
> Hi Shane,
>
> > I was fiddling around with the CVS version of binutils, and noted a
> > bug in one of the macros in gas.
> >
> > .cpload -> s_cpload in binutils/gas/config/tc-mips.c
> > Among other things it maps to a macro_build_lui, (Load upper
> > immediate) which is undefined for MIPS16.  Probably needs an if case
> > for MIPS16, and then use another way to build the memory location
> > up.  (More ops load register or something ?)
>
> I consulted with a MIPS expert about this.  It seems that the .cpload
> directive is only supported by the MIPS32 instruction set, not the
> MIPS16.  Really what should happen is that you get an error message
> saying "unrecognised opcode" or some such.  What actually happens when
> you run into this problem ?

Yes..., I consulted a MIPS expert and they said about the same thing ;-).  
(Said it's too dificult to do the .cpload in MIPS16, which I understand after 
looking at the code)  The error it gives is probably the _true_ bug :

/tmp/ccefNT2B.s: Assembler messages:
/tmp/ccefNT2B.s:18: Internal error!
Assertion failure in macro_build_lui at ../../src/gas/config/tc-mips.c line 
2780.
Please report this bug.

I'm willing to send in a patch against this..., I'll hack it up in a few 
hours.  The assertion that it blows is that it realizes it's in mips16 mode 
at macro_build_lui, but it should have realized this, and blurted out an 
error when it first encountered the .cpload macro in the assembler that was 
put out by the c compiler.

Thanks,
Shane.