Re: VAX addressing modes

[Paul Koning <paulkoning%comcast.net@localhost>]

> On Jul 27, 2017, at 8:56 AM, Martin Husemann <martin%duskware.de@localhost> wrote:
> 
> Ok, here are excerpts from the asm output of gcc:
> 
> .L7814:
>        casel %r0,$0,$7
> .L3234:
>        .word .L3233-.L3234
>        .word .L3232-.L3234
>        .word .L3235-.L3234
>        .word .L3236-.L3234
>        .word .L3237-.L3234
>        .word .L3238-.L3234
>        .word .L3239-.L3234
>        .word .L3240-.L3234
> .L7810:
>        movl 116(%r9),%r0
>        jeql .L3169
> ...
> 
> 
> so this matches your description and makes sense. All .word were of the
> .L... - .L... form, nothing refers to absolute addresses or external
> symbols.
> 
> Now back to the original question: the .word are 32bit offsets relative
> to the casel instructions?
> 
> If so, I'd call this a binutils bug. No relocations should ever be generated.

Not usually, agreed.  But you would get a relocation if the destination is in a different section than table.

	paul