Re: TextAid on the Iyo

In message <934b68e14d.druck@xxxxxxxxxxxxxxxx>
druck <news@xxxxxxxxxxxxxxxx> wrote:

> On 30 Dec 2005 Alan Wrigley <spamhater@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
> wrote:
>> It's full of bits of code that either might rely on flags being returned
>> from subroutines, or might rely on flags *not* being returned from
>> subroutines. Armalyzer is no help in these situations
>
> It should be, it has quite a bit of logic to determine if a subroutine is
> preserving or returning flags and if the calling routine is using them, or
> relying on a previous conditions being preserved.
>
> Most code was written as flag preserving as a matter of good programming
> practice, but in most cases this property isn't used by the program, and
> when converted to 32bit can be left flag corrupting.
>
> The exceptions are usually small backs of code which conditionall call
> subroutines or SWIs and rely on the flags being preserved to be used in
> further conditional calls, e.g.:-
>
> CMP something
> BLEQ oneway
> BLNE theotherway

This is precisely one of the problems I had in mind. There's probably
a lot of this in the assembler programs I wrote in the early 90s and
is the main reason why I don't want to 32-bit them. OK, with hindsight
it could have been done differently, but there are plenty of other
13-year-old apps that would work better now if hindsight could be
applied.

Alan

--
RISC OS - you know it makes cents
.

Relevant Pages