forums.silverfrost.com

[brucebowler]

I have some code that has a bit that looks like this

[brucebowler]

[Martin]

Hi Bruce,

I could not reproduce the problem mentioned in your first message. Which version of FTN95 did you try this with?

I am not sure the input file in your second message is valid. You are attempting to specify the values of the chlEnd array, but you only specify the first two values when the array length is 10. FTN95 attempts to read the 3rd value and finds the chlcoeff assignment instead of a digit - so it gives the "Digit(s) expected" error. I'm not sure this is allowed by the standard (although it seems strange if it is only allowed when it is the last assignment in a namelist block). I will check the standard on this and get back to you soon. In the meantime you could explicitly specify the range of the array you are assigning to with:

[brucebowler]

Hi Martin, I'll give that a shot when I get a chance, but I believe it is permitted to give fewer elements than necessary, with the remaining elements filled with "nulls" as needed. The following (I'll admit it's not from "the standard", but it is a fairly authoritative source, at least in my book) confirms that belief.

http://docs.hp.com/en/B3908-90002/ch08s04 says, in part,

"If the name in a name-value pair is that of an array, the number of the values following the equals sign must be separated by value separators and must not exceed the number of elements in the array. If there are fewer values than elements, null values are supplied for the unfilled elements."

[brucebowler]

OK, I had a chance to test it... (this is 5.0.1, by the way).

The problem appears to only occur with 2D arrays if the item immediately prior in the input file is a 1D array that does not have enough data to fill the array.

Add another variable to the namelist called chlStart and dimension it (10) and add that variable to the namelist statement.

make test.opt look like

&optlist
chlstart = 0,1
chlend = 1,2
/

When you run the program, chlStart and chlEnd will have the right values.

Add the chlCoeff line to the end of the input file and it fails with Error 77.

Make test.opt look like

&optlist
chlStart=0,1
chlEnd=1,2,8*0
chlcoeff(...)
/

and the program works.

Make the chlEnd line look like chlEnd(1:2) = 1,2 and the program works.

From this I conclude there's some funky error in the RTL routines that get called when one reads a multidimensional array from namelist.

[Martin]

Thanks Bruce.

The standard does indeed state that fewer values than elements may be specified for arrays, so this is a bug in FTN95.

You are also right in that this is only a problem when a multi-dimensional array is the item immediately following an array that has fewer values than elements. Specifically, it is the "," character within the parenthesis that is throwing it off as it thinks it is a value separator in this case.

I will report the problem within our bug tracking system and we will look into fixing it in the near future.

[DanRRight]

If it is a bug was it fixed? Either i have some corruption due to 2GB limit or i've got similar problem unable to load for two damn nights some simple array with format like the author above.

By the way, what "Digit(s) expected" in Error 77 means?

I gave up after whole weekend day and night digging...things are that this definitely somehow worked before. Preliminary, when i define format explicitly

iz=5
READ (A,'(30x,30(e11.3))') (Var(ii,iz),ii=1,30)

all works fine, but when use character variable as a format:

iz=5
charFormat='(30x,30(e11.3))'
READ (A,charFormat) (Var(ii,iz),ii=1,30)

i get error 77...What is the length limit for charFormat variable i can use?