READ not working

[Brooks_Van_Horn]

I am developing a windows dialog based program and have a problem with:

       num = GetDlgItemText(hwDlg, IDC_M1, myBuffer,20)
       If (num /= 0) Then
          iostatus = 1
          do while (iostatus /= 0)
             Read(myBuffer,*,iostat=iostatus) X
          end do

myBuffer = '25.6                  '
num = 4
and X = garbage
iostatus = 59

without the do while and the iostat, the program has a break at the read statement. When I programmed in C++, there was a lock function to keep myBuffer in real ,memory so the i/o works. This is part of a callback routine. What can I do to fix this?

Brooks

[jimdempseyatthecove]

Are you saying, without the do while, you suspect that between the GetDlgItemText into myBuffer, and the internal READ from the text in myBuffer into the numeric X, that the O/S paging system paged out the memory with myBuffer?

If there is a paging issue, and I emphasize if, then it is possible that only part of myBuffer was touched (and paged in) by GetDlgItemText. If this is the case, then immediately preceeding GetDlgItemText, insert

myBuffer = ' '

Note, if the above causes a fault, then your myBuffer address is corrupted or uninitialized.

Jim Dempsey

[jimdempseyatthecove]

I should have mentioned that paging should not have caused a problem.

Jim Dempsey

[andrew_4619]

Two points. ISTATUS=0  OK read /= 0 is a read error. You logic this seems strange! Secondly what is the value of IOSTAT after the first read, that this give you the indication of why it failed.

[jimdempseyatthecove]

Andrew,

he is setting IOSTATUS=1 prior to the do while loop. He expects it to iterate once, but is stating he is experiencing it iterating more than once. His suspicion is that myBuffer experiences a page fault, and that this fault returns an error (status) that is non-zero. Then the second time through, no page fault, no error. Page faults should not be returned as an error (except for non-pagable interrupt level code).

GetDlgItemText will not padd the buffer with blanks, it returns the number of valid bytes. Brooks should pre-set the buffer to ' ' prior to call (or use the length returned to restrict the "input" to the valid bytes:

Read(myBuffer(1:num),*,iostat=iostatus) X

Jim Dempsey

[andrew_4619]

D'oh, I should have read more closely. But 59 is a read format error so the buffer string presumably has some junk in it that cannot translate to a number? Is buffer actually declared as 20 long? I normally pass len(buffer) BTW.

[Steven_L_Intel1]

It's always difficult to diagnose a problem when all we have is snippets, descriptions and pseudo code, but I am going to guess that the callback routine does not include a !DEC$ ATTRIBUTES STDCALL directive, leading to stack corruption. (Note that STDCALL by itself will change to pass-by-value, so add REFERENCE if you want that.) Windows dialog callback routines must be STDCALL. I don't think page faults have anything to do with it.

[GVautier]

jimdempseyatthecove wrote:

Read(myBuffer(1:num),*,iostat=iostatus) X

Effectively, that's the way the read from the string returned by GetDlgItemText must be done.

Fix your code before searching for another error