|
|
I got this old Fortran exe and I have to embed in in C++. After starting it needs two filenames entered by the user. I did this by putting the filenames in files, attaching handles to them, making a Pipe and attaching it to the child process, then doing ReadFile and WriteFile to the pipe, as shown in MSDN.
Anyway, it gets the strings holding the filenames. I have to append \n else Fortran won't accept it. The Fortran people advised me to try with \r before \n, it seems to grab that too. C++ can append a \0 if I let it, I tried that too.
However, when it takes the first string and tries to open the file, it can't do it and it returns an error. (It returns the error tidily to my C++ process, so it can't be that bad.)
If I run it directly from the command line, as the original author did, it all works fine with the same files. So there must be something wrong with the way I pass my strings.
Unfortunately I can see into the Fortran code but I can't get at it to see what the problem is. We have some old Fortran compilers lying about, but they won't even compile it. The original came from Linux, can that have something to do with it? The users have real problems if they use a Windows editor, it doesn't like Windows carriage returns, they say. But I run it from the command line and it reads all kinds of other files I generated with C++.
There must be someone out there who has done this before?
-------------
Bibo ergo sum
|
|
|
|
|
If it runs OK from the command-line, why don't you pass in the filenames on the command-line when you call CreateProcess() ?
|
|
|
|
|
It runs fine if I type in first "fortran.exe 1>test.out 2>test.log", then the two filenames "input.dat" and "test", both followed by CR. The Fortran.exe is waiting for two inputs on UNIT=5. My problem was to embed all this in my real-time stuff, so it gets called without the user realising. As far as I know that bit works, it does continue after it is fooled into believing it has 2 keyboard inputs. It just doesn't like the inputs.
Did I go to all that trouble when there was an easier way of doing it? I couldn't even get a batch file to do it.
-------------
Bibo ergo sum
|
|
|
|
|
Oh, I didn't realise the fortran.exe program gets the input for the two filenames.
|
|
|
|
|
Hi,
have you tried passing the Fortran exe filenames ending on \r (i.e. without any \n)?
can you determine whether it is the first or the second file it fails to open? I suspect it
understands only one line terminator, and the other one (if you use \r AND \n) is left unrecognized
in the other filename.
|
|
|
|
|
So many answers, such kind people... This is what I came for.
Anyway, yes, I tried just \r but these were not recognised as strings i.e. IOSTAT was not 0.
READ(5,'(A)',IOSTAT=IOS) FILEN
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A FILE NAME'
READ(5,'(A)',IOSTAT=IOS) FLBASE
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A ROOT FILE NAME' *** it complained here
Now, with \n or with \r\n it takes the strings and continues. It is the first file it fails to open. I wonder why it complains at the second input and not at the first.
Maybe I should input both in the same string... You may have something there with the line terminators... Shall try that.
-------------
Bibo ergo sum
|
|
|
|
|
|
I tried to look at the error code but it is a little smiley in my command window. I have to find out how to format Fortran to write in hex, oh dear.
-------------
Bibo ergo sum
|
|
|
|
|
Hi,
I would try "filename1\rfilename2\r" as a single string; I suspect \r is the right separator
for Fortran, NULL is needed on the sender side, and an extra NULL in between the filenames
ends up in FLBASE[1].
If the final NULL is confusing something else later on, there are ways to send a string without
a terminating NULL too.
A final thought:
I understand you can no longer compile the full Fortran app.
If your tests fail to show the way, you could write just a small Fortran program that tests
the communication and shows all the characters that actually get read into FILEN and FLBASE.
|
|
|
|
|
No, putting them in one string didn't work, it really wants two, with \n. It doesn't seem to care about C++'s \0. Does it read them out of the pipe in the same order as I put them in? I tried both ways but it didn't work either.
That is a good idea with the little test program. I have just written my first Fortran code in 20 years. Ugh, they call that a debugger? I shall now try and run that from C++. Progress report to follow...
-------------
Bibo ergo sum
|
|
|
|
|
What happened? When I call my little ftest.exe from the debugger or from the command window it works, I have to enter the two strings by hand. I can see that it has filled the input strings up to 80 chars, but still manages to open the file.
When I call it from C++ it fills the strings up to 80 chars, but can't open the file. I tried terminating with \n, with \r\n, and with \r\n\0, it all made no difference.
There must be something else. This is very strange.
READ(5,'(A)',IOSTAT=IOS) FILEN
It must be the '(A)'
-------------
Bibo ergo sum
modified on Tuesday, December 9, 2008 8:22 AM
|
|
|
|
|
Hi,
AFAIK '(A)' just means "take any text"
I suggest you:
- make sure your debugging stuff visualizes all special chars including space, NULL, CR, LF
- display the status of both openfile operations, I mean the actual value so you can google it
(a >0 test is not sufficient)
- try the Fortran test app manually with different paths
- show us all the relevant Fortran code
|
|
|
|
|
This is the entire code:
PROGRAM FTEST
IMPLICIT NONE
C
INTEGER IOS
CHARACTER*80 FILEN, FLBASE
C
C
CONTINUE
READ(5,'(A)',IOSTAT=IOS) FILEN
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A FILE NAME'
WRITE (6) 'FILEN = *', FILEN, '*'
READ(5,'(A)',IOSTAT=IOS) FLBASE
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A ROOT FILE NAME'
WRITE (6) 'FLBASE = *', FLBASE, '*'
OPEN(17,FILE=FILEN,FORM='FORMATTED',STATUS='OLD',
$ READONLY,IOSTAT=IOS)
IF ( IOS .NE. 0 ) THEN
WRITE (6, FMT=*) 'ERROR NO INPUT FILE ', IOS
ELSE
WRITE (6) 'OPENED INPUT FILE OK'
ENDIF
C
CONTINUE
END
-------------
Bibo ergo sum
modified on Tuesday, December 9, 2008 8:46 AM
|
|
|
|
|
OK, now
1. what is the result opening the first file? is it ERROR ... or OK on first file?
2. where is the opening of the second file? and what is its result?
|
|
|
|
|
It gives me an IOS=2 on the first file. I will worry about the second string another day, it gets used to build some complicated output filenames.
-------------
Bibo ergo sum
|
|
|
|
|
Hi,
Not sure but maybe IOS=2 means "the file you want to read exists but is empty". Please check.
I expect IOS=6 for "no such file", IOS=0 for anything else (until you start working on the file that is) unless the filename is invalid (not sure which code gets emited then).
You could try creating and writing the file, rather than opening and reading; that way, Explorer (or DOS DIR) would show the effective filename.
|
|
|
|
|
Hi,
I installed Silverfrost F95 and was able to run this:
PROGRAM FTEST
IMPLICIT INTEGER(I-N)
INTEGER IOS
CHARACTER*80 FILEN, FLBASE
CHARACTER C
PRINT *,'enter filename1'
READ(5,'(A)',IOSTAT=IOS) FILEN
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A FILE NAME'
PRINT *, 'FILEN = *', FILEN, '*'
DO 10 I = 1,80
C=FILEN(I:I)
IF (ICHAR(C).NE.32) PRINT *, 'FILEN[',I,']=',C,'=',ICHAR(C)
10 CONTINUE
PRINT *,'enter filename2'
READ(5,'(A)',IOSTAT=IOS) FLBASE
IF ( IOS .NE. 0 ) STOP 'MUST PROVIDE A ROOT FILE NAME'
PRINT *, 'FLBASE = *', FLBASE, '*'
DO 20 I = 1,80
C=FLBASE(I:I)
IF (ICHAR(C).NE.32) PRINT *, 'FLBASE[',I,']=',C,'=',ICHAR(C)
20 CONTINUE
OPEN(17,FILE=FILEN,FORM='FORMATTED',STATUS='OLD',IOSTAT=IOS)
IF ( IOS .NE. 0 ) THEN
PRINT *, 'ERROR NO INPUT FILE 1 ', IOS
ELSE
PRINT *, 'OPENED INPUT FILE 1 OK'
ENDIF
OPEN(18,FILE=FLBASE,FORM='FORMATTED',STATUS='OLD',IOSTAT=IOS)
IF ( IOS .NE. 0 ) THEN
PRINT *, 'ERROR NO INPUT FILE 2 ', IOS
ELSE
PRINT *, 'OPENED INPUT FILE 2 OK'
ENDIF
END
it gives OK for existing files, and error 128 for non-existing files.
The documentation says error codes could be different on different compilers.
So it is my guess something is wrong in the way you pass the filenames from C++;
anyway the above should be able to tell you what gets read.
|
|
|
|
|
Luc Pattyn wrote: I installed Silverfrost F95
nice one Luc - thats dedication !!!
'g'
|
|
|
|
|
My very first large apps were coded in Fortran IV, some 30 years ago, but I hadn't done any Fortran work since I switched to C, also long ago. This was an opportunity to read some of the docs, and experiment a bit.
|
|
|
|
|
Thankyou. You have helped me immensely. It almost works now; I can run my test exe from C++, now "just" need to get the real thing running.
If I discover anything new I will report it. Meanwhile I am very grateful.
-------------
Bibo ergo sum
|
|
|
|
|
Another possible difference to running it from the command line is the location of the working directory (I note that you seem to be using filenames and not full paths). Maybe you are now successfully transferring the filenames but the working directory is different, perhaps somewhere where the program doesn't have permission, or a file that should exist in it doesn't.
The statement:
OPEN(17,FILE=FILEN,FORM='FORMATTED',STATUS='OLD',
$ READONLY,IOSTAT=IOS)
will give you some sort of error if the file doesn't exist as STATUS='OLD' tells the program that it should be there.
Peter
"Until the invention of the computer, the machine gun was the device that enabled humans to make the most mistakes in the smallest amount of time."
|
|
|
|
|
Yes yes yes!
I put the whole path in (without doubling the \) and the test program works. It doesn't work for the original Fortran yet but at least I know where to start. If Fortran is only going to give it 80 chars every time I can see some problems coming up on the target machine, because the paths can get a lot longer. I did try a relative path but it wasn't having it, must try and get Fortran to tell me where it is. In C++ I use _getcwd, can someone tell me how to get it in Fortran?
It was giving me IOS=2, which is not in the list of Intel error messages.
-------------
Bibo ergo sum
|
|
|
|
|
To find out where you are, GNU Fortran has a GETCWD call
link[^] I don't know how standard it is.
It doesn't seem to have a corresponding SETCWD call.
If you are starting the program using CreateProcess() then you can set the working directory, and I think it will default to the working directory of the calling process if you don't.
Peter
"Until the invention of the computer, the machine gun was the device that enabled humans to make the most mistakes in the smallest amount of time."
|
|
|
|
|
Unfortunately the Absoft compiler is a bit basic and does not have such luxuries. But I must check what I put in CreateProcess. I have to put it all in the command line, maybe I missed the path and it was just luck that it ran where it is.
-------------
Bibo ergo sum
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
