Another "debug" vs. "release"

[lklawrie1]

The two configurations seem to differ. The release is flat out crashing. The debug runs. Making it very hard to track down if it's a real error or not.

CVF works fine. (At least in debug mode, I suppose I can try the release mode there too).

If I can't resolve this soon, I will have to switch compilers to ???

[Steven_L_Intel1]

What I usually do in these cases is, in the debug configuration, change the optimization level and see if the problem returns. I then selectively lower optimization on source files until I find one that makes a difference. Sometimes you can lower the level partially (to 2 or 1) and still see the error, yet still have good debug information.

[lklawrie1]

Yes, that will work if it's truly a problem in the code. But if it's the compiler configuration? Ref: earlier support issue. Is anyone working on that other than you?

Linda

[lklawrie1]

Ok, you're off the hook this time -- currently works iff there is a SAVE on a var in the calling routine -- now I have to figure if that is accurate.

And, it won't compile right now in Lahey anyway. (Which will have to change when I do the Linux port).

Linda

[Steven_L_Intel1]

That can be a sign of an uninitialized variable.

[lklawrie1]

Might be uninitialized.

I'm still getting the NANs (which was the next reason this one died since I had commented out that code for trying the Lahey compiler). Any idea if they're related to the optimization level?

Linda

[Steven_L_Intel1]

Impossible to say. Usually, problems that show up with optimization are due to incorrect assumptions made by the source code. Optimized code is less forgiving of coding errors. But there could also be a compiler bug.