#1750 INTERPRET .context~digits different from digits()

[jfaucher]

When some numeric settings are set inside an INTERPRET, and queried from the same INTERPRET with RexxContext ~digits or ~fuzz or ~form then the results are different from digits() or fuzz() or form().

The attached file displays 3 lines. The second line shows the problem.
The last number is the RexxContext~identityHash

15 14 ENGINEERING 15 14 ENGINEERING -4406334017
15 14 ENGINEERING 9 0 SCIENTIFIC -4406334017
15 14 ENGINEERING 15 14 ENGINEERING -4406334017

This problem occurs with ooRexx 4.2 and 5.0.
A possible fix is described in the attached file.

[jfaucher]

The proposed fix is not working when passing the current context to a routine or method.
Attached file updated with this new test case (v1).

Output (without proposed fix) is ok:
Caller context: 15 14 ENGINEERING -4528229313
Current context: 100 0 SCIENTIFIC -4528283569
Caller context: 15 14 ENGINEERING -4528229313

If the proposed fix is applied, the output is KO:
Caller context: 9 0 SCIENTIFIC 567214664
Current context: 100 0 SCIENTIFIC 567220084
Caller context: 100 0 SCIENTIFIC 567214664

Maybe the instruction numeric digits should update both the current RexxActivation and its parent RexxActivation when isInterpret() == true ?

[jfaucher]

After testing the 2nd proposed fix, it appears that it should be like that to support nested INTERPRET:
recursively updating the setting while isInterpret() == true
This is the same logic as parent->rexxVariable(name) in RexxActivation::rexxVariable.

void RexxActivation::setDigits(wholenumber_t digitsVal)
{
settings.packageSettings.setDigits(digitsVal);
if (isInterpret()) parent->setDigits(digitsVal);
}

Attached file updated with a new test case for nested INTERPRET (v2).

[Erich]

Fixed code and added test with revision [r12187]