Gpf

[Guest]

Hi all,

Can someone please explain why this code gpf?
Not always...just sometimes. I double checked
not going beyond array bonds and the pointers > 0,
...same thing. I'm not using strptr(string) but varptr(string).
I isolated this offending snippet. Actually, the type shown
below has many other pointers, but just the string ptr gpf.
The weird thing, it sometimes gpf when reading, others when
trying to assign a string. There is something I'm doing wrong?

Thanks in advance

Gus

type main_type
   nstr as string ptr
   idx as long
end type

function mainthing() as long     '<- main entrance
  local mtp as main_type
  local pmtp as dword
  local names() as string
  redim names(512)
  mtp.nstr = varptr(names(0))
  mtp.idx = 3
  pmtp = varptr(mtp)
  
  
' then some processing...ie
  redim preserve names(ubound(names)+1)
  array insert names(), "this thing"
  '...
  names(28) = "this other guy"

' then a call to do some other processing
  dothings(pmst)
  '...
  '...
    
end function

function dothings(byval mtp as main_type ptr) as long
  local st as string
  st = "damngpf"
  @[email protected][@mtp.idx] = st   '<---THIS GPFs
  st =  @[email protected][@mtp.idx]  '<---THIS also GPFs

  @[email protected][1] = st   '<---THIS GPFs
  st =  @[email protected][1]  '<---THIS also GPFs!!!

end function

[Kev Peel]

By glancing at the code I see that your REDIM PRESERVE and ARRAY INSERT statements may be reallocating the array data, making the previously obtained pointer invalid. Try obtaining the names(0) pointer after the REDIM, to remedy.

[Michael Mattias]

ARRAY INSERT won't change the address of element zero, but REDIM will for sure.

i.e, Kev's fix will work.

[Guest]

You right.

It is not supossed string handles doesn't moves?

Thank you.

[Michael Mattias]

You right.

It is not supossed string handles doesn't moves?

Thank you.

REDIM causes the block of memory allocated to hold the array elements to move, so the address of the string handle (the "STRING PTR") changes on REDIM.

MCM

[Guest]

I know redim moves the actual string data, but I'm using VARPTR and supossed (evidently wrong) handles was fixed.
Anyway, just refreshing the pointers after any redim did the trick.
This thing flyes now

Thank you both.

[Michael Mattias]

I know redim moves the actual string data, but I'm using VARPTR and supossed (evidently wrong) handles was fixed.
Anyway, just refreshing the pointers after any redim did the trick.

Actually, the string *data* do not move at all during a REDIM PRESERVE of a STRING array...i.e, the STRPTR would still be valid. In this case the STRPTR is useless (you need the VARPTR) but it would remain valid.

(I think that's correct: it wouldn't make sense to rebuild all the strings when not necessary).

But since it's working correctly now, do you really care?

MCM

[Bob Zale]

In a REDIM PRESERVE:

1- The array descriptor never moves.
2- The string data never moves.
3- The string handles which point to the string data never change.
4- The block of memory where the string handles are stored may change if the block can not be expanded to the new size.
5- VARPTR() of an array element may change.
6- STRPTR() of an array element will not change.

Bob Zale
PowerBASIC Inc.

[Arthur Gomide]

Bob,

I think it would be very helpful to include these comments on the clause "REDIM PRESERVE" in the Help file.

Best Regards,