Re: [Gm2] StringConvert dialect

[Gaius Mulley]

John B Wallace Jr <address@hidden> writes:

> There seems to be a slight problem with the StringConvert module. It appears
> that it is written in the PIM2/PIM3 dialect and when compiled with the default
> dialect of the most recent gm2 (PIM4 I believe), it produces some errors.
> 
> 
> The two test modules below produce an incorrect result for MIN(INTEGER) and
> MIN(LONGINT) when built with the 16Sep05 version of gm2:
> 
> 
> MODULE InOutBug;
> 
> (*******************************************************************)
> (*        NOTE: THIS IS TEST CODE AND MAY BE INCORRECT             *)
> (*******************************************************************)
> 
> FROM InOut IMPORT WriteInt, WriteLn;
> 
> BEGIN
>   WriteInt(MAX(INTEGER),0);
>   WriteLn;
>   WriteInt(MIN(INTEGER),0);
>   WriteLn;
>   WriteInt(MIN(INTEGER)+1,0);
>   WriteLn
> END InOutBug.
> 
> The output is:
> 
> 2147483647
> -2147483658
> -2147483647
> 
> 
> MODULE FpuIOBug;
> 
> (*******************************************************************)
> (*        NOTE: THIS IS TEST CODE AND MAY BE INCORRECT             *)
> (*******************************************************************)
> 
> FROM StrIO IMPORT WriteLn;
> 
> FROM FpuIO IMPORT WriteLongInt;
> 
> BEGIN
>   WriteLongInt(MAX(LONGINT),0);
>   WriteLn;
>   WriteLongInt(MIN(LONGINT),0);
>   WriteLn;
>   WriteLongInt(MIN(LONGINT)+1,0);
>   WriteLn
> END FpuIOBug.
> 
> The output is:
> 
> +9223372036854775807
> -9223372036854775818
> -9223372036854775807
> 
> 
> I was able to correct the problem by compiling the StringConvert
> module with the -Wpim2/-Wpim3 switch or by changing the StringConvert
> code very slightly, replacing a few DIV with / and a few MOD with REM
> in the first part of the IntegerToString and LongIntegerToString procedures.
> 
> The NumberIO module may have the same problem.

Hi John,

many thanks for reporting these bugs - they have all been fixed now
and checked into the current CVS together with runtime regression
tests. As you suggested there was a similar bug in NumberIO
which has also been fixed,

regards,
Gaius