internal error 2014062901
lazarus / 64-FPC development

----------------------------------------------------------------------
engkin · 2021-10-03 06:24

The following code produces an internal error on Mac Big Sur M1 using
trunk:
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i program
project1; {$mode objfpc}{$H+} function bug(s: string): string;var 
p:pointer=@s;//internal error 2014062901beginend; beginend.
Can you please check different OS/FPC version before I report it?

----------------------------------------------------------------------
bytebites · 2021-10-03 06:44

Same result on Linux 3.2 and trunk

Do you expect this result, which prints true
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i function
bug(s: string): string;var  //p:pointer=@s;//internal error
2014062901  p:pointer absolute s;begin  writeln(string(p)); 
writeln(pchar(s)=p);end;

----------------------------------------------------------------------
engkin · 2021-10-03 06:59

Thank you for testing.

Yes, I do expect True in this case. The record holding the length, ref
count, code page ... is at a negative offset of s. Or did I miss your
point?

----------------------------------------------------------------------
bytebites · 2021-10-03 07:32

Length like so
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i  
writeln(PSizeInt(p-sizeof(sizeint))^=length(s));

----------------------------------------------------------------------
Jonas Maebe · 2021-10-03 13:05

While it shouldn't cause an internal error, it is an invalid
expression. The reason is that @s is not constant (it's the address on
the stack of the ansistring, which can be different every time the
function gets called). At least FPC has no support for initialising
variables with non-constant expressions.

----------------------------------------------------------------------
engkin · 2021-10-03 17:52

Yes, it is a faulty code. Just the internal error should not happen.

Reported as issue #39393