Control edit descriptor: $ in a FORMAT

[john_folwell]

When I compiled an old piece of code on 9.0 I was getting strange behavior when I tried to run it. Somtimes it just quit and went away in the middle of the process. Sometimes it gave me a: severe(157): Program Exception - access violation. To make a long story short, when a $ is used to output information from two write statmentsto one line, the program goes down the tube. I'm fairly sure this worked on 8.1 Here is a short code sample that fails when run as a qwin application.

c jqwin1 - illustrate '$' problem from legacy code
c 09/30/05 jdf originate
program jqwin1
write(*,'(a,$)')'The quick brown fox jump'
write(*,'(a)')'ed over the lazy dogs.'
end program jqwin1

[Steven_L_Intel1]

This works fine for me in 9.0.021, and I haven't run across a report of this error before. What compiler version are you using?

[john_folwell]

I am running 9.0.021 on a Dell 5150.

[Steven_L_Intel1]

Please submit a report to Intel Premier Support and attach a ZIP of a sample project that shows the problem. I could not reproduce this on my own.

[kmervin]

I am getting similar behavior in an large old program we are converting from

CVF to IF 9.021. I do indeed have a lot of $ formats, but I'm not necessarily

convinced that's the issue.

I could not reproduce the short test program error

either. Where exactly my program crashes seems to depend on the machine, Debug or not, and optimization settings, so it's tough to pin down.

If you find a solution, please post back her.

Kent

[john_folwell]

To kmervin: The program I was working on was not all that large (900 lines of code). There was nothing to connect the problem with $. When I said long story I meant it. When all the $'s were removed all the problems went away

[Steven_L_Intel1]

Please submit an issue with what test case you have. I'll also discuss this with the appropriate engineers.

[adam_bird]

This error appeared to occur randomly and was very hard to re-produce.
The error was being generated by a WRITE statement writing to a recently opened scratch file. The actual cause of the error was an overwrite a few lines of code earlier.
Like the first message this error is in code that has been used by clients for years with no problem and works fine with other compilers and OSs.
Good luck looking...

[emc-nyc]

Back in the day, some compilers used a $ to suppress line feeds in display output.


Just replace



write(*,'(a,$)')') 'this is an ugly prompt:::'


with



write(*,'(a)',advance='NO') 'this is an ugly prompt:::'





added in edit





I'm sorry; I completely missed the gist of this question, which seems more concerned with something which may -- or may not -- be going on in QuickWin or Intel's i/o system. Everytime (twice) I've tried to use write(*,'()') in quickwin i've gotten the same error, so I cannot be of help.

Message Edited by emc-nyc on 09-22-200510:43 AM

Message Edited by emc-nyc on 09-22-2005 10:44 AM

[Steven_L_Intel1]

I had missed that the original problem was a QuickWin application. That I can reproduce. I'd encourage you to report it, though I will also do so. If you report it, please include a reference to T59139 so that the support engineer can associate it with the original.

Message Edited by sblionel on 09-22-2005 12:47 PM

[Steven_L_Intel1]

I just learned that we had a general problem with the QuickWin library for the past few releases that could result in "strange" behavior. I verified that the test case involving $ format is now fixed - the fix will appear in the next update. I'd suggest that others with QuickWin issues try that update when it appears and see if it helps. If not, please report the problem to Intel Premier Support.

[kmervin]

Steve,

Could you please tell us the timing prognosis for this new update? Which previous update was free of the QuickWin issues?

Thanks. -Kent

[Steven_L_Intel1]

I think the initial 9.0 version, 9.0.018, was ok in this regard. According to our schedule, the next 9.0 update is late October, but it's possible one would be available sooner. You could try just replacing libifqwin.lib with the one from that version and see what happens.

If this is a serious problem for you, I would advise submitting an Intel Premier Support issue.

[Steven_L_Intel1]

9.0.024, if you should ever succeed in downloading it, contains the fix for the QuickWin problem.

[kmervin]

9.0.024 fixed our access violation problems, thanks.

[john_folwell]

version 9.0.024 fixed all my current problems thankyou