Re: .asciz - implementation or documentation wrong?

["Jan Beulich" <JBeulich at novell dot com>]

>>> On 06.07.11 at 10:21, Erik Christiansen <dvalin@internode.on.net> wrote:
> On 05.07.11 12:38, Jan Beulich wrote:
>> Documentation as well as the comment prior to stringer() say that the
>> directive (like .ascii, but there it doesn't matter that much) expects
>> comma separated double quoted strings. Yet a missing comma between
>> two strings is taken as if there was one (whereas simply concatenating
>> such strings - without any NUL character in between - would seem like
>> the much better alternative).
> 
> Jan, in comparison, I'm having trouble seeing any documentation vs
> behaviour difference, on a quick check with:
> 
> $ avr-as --version
> GNU assembler (GNU Binutils) 2.20
> 
> Here, input of:
> 
>    .data  
>    .asciz   "AAAA", "BBBB"
>    .asciz   "CCCC"  "DDDD"

This line ought to be rejected according to the documentation.

>    .ascii   "EEEE", "FFFF"
>    .ascii   "GGGG"  "HHHH"

As should this one.

> gives:
> 
>    8                    .data
>    9 0000 4141 4141     .asciz   "AAAA", "BBBB"
>    9      0042 4242 
>    9      4200 
>   10 000a 4343 4343     .asciz   "CCCC"  "DDDD"
>   10      0044 4444 
>   10      4400 

That's precisely what I'm seeing.

>   11 0014 4545 4545     .ascii   "EEEE", "FFFF"
>   11      4646 4646 
>   12 001c 4747 4747     .ascii   "GGGG"  "HHHH"
>   12      4848 4848 
> 
> AFAICS, ascii and asciz are both indifferent to the comma, and
> everything is concatenated as expected, with NULs where required.
> 
> If an inconsistency exists in newer binutils, and documentation does not
> mandate it, then that is a regression, I submit.
> 
> If the post is a proposal to implement, and document, "Only one NUL, if
> no commas", then the breaking of a few existing assembler programs may
> be the price.

I realize that, which is why I didn't send a patch right away.

> The introduction of Awk-like string concatenation syntax
> seems cute, but doesn't achieve anything that can't be done by eliding
> the two quotes and space to begin with.

Eliding the quotes may not be possible - just consider the case of
passing a .S file through a C pre-processor, and the intended string
consisting of multiple pieces, some or all of which are pre-processor
constants.

Jan

> At least then, the program is
> more honestly explicit: z means each string is zero terminated. 
> 
> i.e.:
> "`.asciz' is just like `.ascii', but each string is followed by a zero
> byte."
> 
> So is there any gain to offset the pain of changing that?
> 
> Erik
> 
> -- 
> In a time of drastic change it is the learners who survive; the
> 'learned' find themselves fully equipped to live in a world that no
> longer exists.                                     -  Eric Hoffer