FPC 3.3 trunk error "Operation + not supported for PWideChar and
PWideChar"
lazarus / 62-General

----------------------------------------------------------------------
AlexTP · 2020-08-22 14:22

Quoteatsynedit_regexpr.pas(4759,18) Error: Operation "+" not supported
for types "PWideChar" and "PWideChar"
Code was
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          
if not MatchOneCharCategory(scan + REOpSz + RENextOffSz, reginput)
then Exit; I have only 1 PWideChar (scan) and others 2 are offsets
(Int/Int64).
Full repo https://github.com/alexey-t/atsynedit

FPC 3.2.1 didn't give this error. Linux x64.
Free Pascal Compiler version 3.3.1-r46541 [2020/08/22] for x86_64

----------------------------------------------------------------------
PascalDragon · 2020-08-22 15:34

Please report a bug with the attached code as an example so that it
isn't forgotten.

The problem appears to be related to inlining, so as a workaround you
can disable the inline directive for the MatchOneCharCategory function
for 3.3.1.

Edit: just saw that you reported it already. I've attached the example
there.

----------------------------------------------------------------------
Zvoni · 2020-08-25 10:23

Maybe Related?
https://forum.lazarus.freepascal.org/index.php/topic,44713.msg314696.h
tml#msg314696

----------------------------------------------------------------------
PascalDragon · 2020-08-25 11:40

Quote from: Zvoni on August 25, 2020, 12:23:02 pmMaybe Related?
https://forum.lazarus.freepascal.org/index.php/topic,44713.msg314696.h
tml#msg314696

No. Your problem (did you report it?) always triggers in that
situation, however the bug here is only visible if the function is
inlined so it's something where the optimizations of the compiler make
a mistake.

----------------------------------------------------------------------
Zvoni · 2020-08-25 11:50

Quote from: PascalDragon on August 25, 2020, 01:40:08 pm
No. Your problem (did you report it?) always triggers in that
situation, however the bug here is only visible if the function is
inlined so it's something where the optimizations of the compiler make
a mistake.
Thaddy reported it.
https://bugs.freepascal.org/view.php?id=35248