[SOLVED] When is nil not nil?

[BigChimp]

x86 FPC trunk, Lazarus trunk:
I'm feeling fairly stupid, but... here goes:
I had code like this (dbf_dbffile in FPC trunk):

Code: [Select]

procedure TDbfFile.SetFieldData(Column: Integer; DataType: TFieldType;
  Src, Dst: Pointer; NativeFormat: boolean); 
...
var
...
  Len: Integer;


Code: [Select]

          Len := PWord(Src)^;
          if Len > FieldSize then
            Len := FieldSize;

I suspected Src could be an unassigned pointer and changed the code to

Code: [Select]

        if Src = nil then
          Len := 0
        else
        begin
          Len := PWord(Src)^;
          if Len > FieldSize then
            Len := FieldSize;
        end;

Running through the debugger, I break on the first line, if Src = nil.

The debugger shows Src is indeed nil, but when I try to step past, I get a SIGSEGV.

Any ideas?

[Bart]

Probably a debugger (gdb) problem?
Throw in some writeln()'s to debug it?

Bart

[BigChimp]

Edit: not solved yet:
I had been testing dbtestframework_gui against MS SQL server under Lazarus debugger.

When I compile & run dbtestframework.pas, the command line version, edit: using the SAME db as previously (instead of a different one that does not support the field type in question... sorry), the test still has errors:

    00.162  TestDBFExport_VisualFoxPro  Error: EAccessViolation
      Exception:   Access violation
      Source unit:
      Method name:
      Line number: 0

Now going to add some writelns...

[BigChimp]

Interesting. On my Linux x64 system, the second code snippet works fine. It should be running the same tests (same database backend) unless I messed up again.

Throwing in some writelns and recompiling indicates the text mode test program works fine on Windows now, too, so I guess it was the debugger in Lazarus after all....

Thanks.