On 2017-03-05, at 14:40, Martin Truebner wrote:
>
>>> ... to have heard from either of the two commerical vendors of
>>> HLASM compatible cross-assemblers.
>
> Now I have seen one. What is the other (is there one?) ?
>
Dignus, Tachyon, (does MicroFocus have one?), and the non-commercial
z390.
A while back, I discovered that:
X EQU X'7FFFFFFF'
USING R2,X
LA R3,-X
Resolves to the equivalent of:
LA R3,2(,R2)
... where I believe it should say:
ASMA307E No active USING for operand 2
I surmise it improperly ignored an overflow as it tried to resolve.
This matters little for AMODE 24 or AMODE 31; much for AMODE 64.
Does this behavior persist? Cross-assemblers?
On 2017-03-05, at 12:31, Tony Harminc wrote:
>
> Do you think 64-bit expressions are really what's needed, or should it
> be infinite-precision expressions with the ability to
> truncate/round/etc. results explicitly (and implicitly based on
> operands)? I don't know the HLASM internals (thanks to OCO after
> ASMH), but I imagine implementing infinite precision wouldn't be a
> huge amount more work than hardcoding 64-bit. And certainly 128-bit is
> coming from behind.
>
Tread carefully to avoid either massive incompatibility with existing
code or serious compromise of current error checking.
The most plausible argument I have heard against 64-bit expressions is
incompatibility with ASMADATA format.
-- gil
