debug help!

debug help!

I have been using the CVF debugger for a long time now. I am now using version 6.6A. As in this version and previous versions, i can't seem to force it to do what i want. I am getting a NaN in some of my variables. well, to trace it back is very difficult. if the debugger would simply raise an error when this condition arises, that would solve my problem. Ie, as soon as there is a divide by 0 or other situation to cause a NaN to occur, i want to see it. Is there a debug option to set? I have looked and looked, but have not found anything to trigger a break when there is a NaN.

Thank you!

7 帖子 / 0 全新
最新文章
如需更全面地了解编译器优化,请参阅优化注意事项

Related to my last message, i'd like to get the debugger to break into code when there is any usual situation. to me a variable that returns a value of "infinity" is bad. i think this should always be an error condition. NaN is also an error condiction that must be debugged. Please let me know how to flag this event!

Thanks!

Steve Lionel (Intel)的头像

Project..Settings..Fortran..Floating Point..Floating Point Exception Handling..0. This should cause an error to be raised when a NaN would otherwise be generated.

(I tried moving your "continued" question here, but the system lost it instead. Sorry!)

Steve

Steve

I got a response to this message on the "watch", but the message didn't show up here. you said to put the floating point exception handling to 0. this is already done. as a matter of fact, this has always been set on all of my projects and it never helped. on multiple versions of CVF (from 6.1 to 6.6A), i have had problems with this. it never catches my NaN or Infinity. Mostly I am concerned with NaN, but infinity would also be nice. can you tell me what the problem is? i have done this on multiple pc's with various processors, and the same problem has existed.

thanks...

Specify the /fpe:0 switch to get the exception.

James

I take it the forum is striking again, it must be stored on a write only device today. Why vendors don't just use debugged software like standard NNTP servers or mail server type mechanisms to host their discussion groups so as not to torture their participants is beyond me. :-)

James

Steve Lionel (Intel)的头像

Sorry for the forum problems, folks. They've been fixed - at least for now.. (sigh...)

Re the original problem -

I heard from "vader1000" by e-mail - he hadn't said here that his Fortran code is in a DLL called by VB. This changes everything - it requires that you call some exception handling routines to enable fpe:0 functionality, and you'll still miss the Fortran exception handling. This is discussed in a chapter of the Programmer's Guide (it was added after the printed edition) and there's a VBVF1 sample that was added for 6.6.

Steve

Steve

登陆并发表评论。