Exceptions in OS/2

[Allan Mertner]

I have finally got a copy of OS/2 up and running, but don't have any of the truly ancient versions I used to use. Instead, I am using a copy of eComStation 1.1, and it seems to be working fine.

However, something looks a little fishy: All of the OS-level exceptions I get raised are of type EInvalidOp ($C000000097), rather than of type EOverflow, EZeroDivide, etc.

Is this something I have broken, or is it how eComStation works? If one of you have an old copy of OS/2, can you try to run the AllTests example for me please and see whether it works? I have attached an executable (OS/2) copy here, but it would be useful to compare what this one does against what happened with executables made with an earlier version of VP.

Let me know, please :)
Allan

[Yoda]

eComStation doesn't differ from OS/2 afaik..at least not for things that VP/2 would use.
Output for your testapp here on eCS 2.0 rc5:

F:\x-down>AllTests.exe
Running test suite...
- tSysTest - Ok
- tCommandLineTest - Failed (Paramcount)
- tTestSinCos - Ok
- tTestSinCosDeriv - Ok
- tTestTan - Ok
- tTestArcTan - Ok
- tTestIntPower - Ok
- tTestPower - Ok
- tTestExp - Failed ()
- tTestSinh - Failed ()
- tTestCosh - Failed ()
- tTestTanh - Ok
- tTestArctanh - Ok
- tTestRounding - Ok
- tTestHypot - Ok
- tTestCubeRoot - Ok
- tMiscMathTest - Ok
- tTestObjectConstruction - Ok
- String test 1 from FreePascal test suite (Short) - Ok
- String test 1 from FreePascal test suite (Ansi) - Ok
- tTestString4Fpk - Ok
- tTestStringComparisons - Ok
- tTestStringParams - Ok
- tTestCharArrays - Ok
- tTestStringsValStr - Ok
- tTestPChars - Ok
- tTestAnsiStrings - Ok
- tTestBasicExceptions !! Memory leak in bytes: 40
- Ok
- tTestOSExceptions - Failed (Privileged instruction: Unexpected exception "Access violation at address 000107FE. Read of address FFFFFFFF")
-- FAILURE --
Of 29 tests, 5 tests failed:
tCommandLineTest: Paramcount
tTestExp:
tTestSinh:
tTestCosh:
tTestOSExceptions: Privileged instruction: Unexpected exception "Access violation at address 000107FE. Read of address FFFFFFFF"

[Tomas Hajny]

Not sure if it still matters but I actually get slightly different results for OS/2 Warp 4.0 and eCS 1.0 (just the end differs).

 

OS/2 Warp 4.0 FP11:

- tTestOSExceptions - Failed (Float overflow: Unexpected exception "Floating point overflow")

 

eCS 1.0 FP5 (kernel 14.104a_W):

- tTestOSExceptions !! Memory leak in bytes: 68

- Failed (Unexpected exception Invalid floating point operation ())

 

BTW, I'm willing to do other testing, also using older VP/2 versions, etc., but please, e-mail me directly (I guess Allan would have my e-mail address somewhere, or just search for "xhajt03" in Google) because I don't get here very often.