a beginner's segmentation fault

[veselic]

Dear pascallers,
I got
Segmentation fault (core dumped)

to the following simple program

CONST NMAX = 1000;
TYPE NRANGE=1..NMAX;
MATRIX=ARRAY[NRANGE,NRANGE] OF DOUBLE;
var N,M: NRANGE; A: MATRIX;

procedure transp(var n,m:nrange; a: matrix);
var i,j : nrange;
    z : matrix;
begin
for i := 1 to n do
for j := 1 to n do
   z[i,j] := a[j,i];
a := z;
end;
begin
a[1,1] := 1; a[1,2] := 2; a[1,3] := 3;
a[2,1] := 4; a[2,2] := 5; a[2,3] := 6;
transp(n,m,a)
end.

Thanks in advance for help, K. Veselic

[Bart]

You do not initialize n and m, and you don't use m in transp.
(Well, in this example, since the are global variables to the main program, they will be zero. This wil reduce the transp procedure to a := z, since none of the loops will be executed.)

Also, the z := a assigment does not what you seem to expect, since a is passed as a value parameter instead of a var parameter. Therefore a is treated as a copy of the original a, and all changes are made to that copy.
So, after the transp() procedure a will remain unchanged.

Bart

[Bart]

Notice that the NumLib package has code to transpose a matrix of floats.

Code: [Select]

uses
  typ, omv;
var
  A: array[1..m, 1..n] of ArbFloat;  //type ArbFloats is defined in unit typ and it defaults to be double IIRC.
  B: array[1..n, 1..m] of ArbFloat;
begin
  ... some code to fill A
  omvtrm(
    A[1,1], m, n, n,
    B[1,1], m
  );
end.

Bart

[alpine]

IMHO the MATRIX is quite big (1M*SizeOf(Double)) to be placed into the stack twice. 

[Thaddy]

I only get the segmentationfault if I compile for 16 bit.... in TP mode.