[BUG]Another error after svn update on bgrabitmap

[freeman35]

fpcsrc At revision 23430.
lazarus At revision 39876.

Compiling bgrabitmaptypes.pas
bgrabitmaptypes.pas(2433,29) Note: Local variable "errPos" is assigned but never used   -- This line is good idea for clean I think
bgrabitmaptypes.pas(2702,1) Fatal: Internal error 2012090607

[circular]

Hmm, try removing optimizations in compiler options.

[freeman35]

Thank you my friend now compiled but another error :S

/lazpaint-code/bgrabitmap/bgracanvas.pas(281,11) Error: identifier idents no member "StrikeTrough"

in StrikeThrough FPCanvas file line 105
"property StrikeThrough : boolean index 8 read GetFlags write SetFlags;"

so change in bgrabitmap
/lazpaint-code/bgrabitmap/bgracanvas.pas
//   if cf.StrikeTrough then Style += [fsStrikeOut]; old
if cf.StrikeThrough then Style += [fsStrikeOut]; // new just "h" added

[BigChimp]

Freeman is right; there has been a typo fix in FPC recently. IMO the fix should have been adding the correct spelling and marking the old one deprecated but for some reason that wasn't done.
http://www.mail-archive.com/lazarus@lists.lazarus.freepascal.org/msg35239.html

Apparently it's been changed in fixed 2.6/upcoming 2.6.2 and trunk/2.7.1.

To properly fix the code, I would use something like (air code not compiled or tested):

Code: [Select]

{$IF FPC_FULLVERSION>=20602} //changed in 2.6.2 and 2.7
if cf.StrikeThrough then Style += [fsStrikeOut];
{$ELSE}
if cf.StrikeThrouh then Style += [fsStrikeOut]; 
{$ENDIF}

As for internal compiler errors: AFAIU, these should just never happen. If the code is invalid code, a proper error message should be generated... perhaps circular already knows which error it is though... and I vaguely remember there were some fixes in 2.7 to address incorrect code generation when optimization is on.

[freeman35]

Take care if this code from source 

Code: [Select]

{$IF FPC_FULLVERSION>=20602} //changed in 2.6.2 and 2.7
if cf.StrikeThrough then Style += [fsStrikeOut];
{$ELSE}
if cf.StrikeThrouh then Style += [fsStrikeOut]; 
             ^ 
{$ENDIF}


[BigChimp]

If you mean take care copying and pasting? Yes, of course, that's why I mentioned it's air code/untested etc.

Thanks for the correction though

[freeman35]

Yes I mean that. I was think you edit svn source file, and copy from there and paste here

[circular]

Well, ok i've updated BGRACanvas with this code :

Code: [Select]

{$IF FPC_FULLVERSION>=20602} //changed in 2.6.2 and 2.7   
    if cf.StrikeThrough then Style += [fsStrikeOut];
{$ELSE}
    if cf.StrikeTrough then Style += [fsStrikeOut];
{$ENDIF}


As for internal compiler errors: AFAIU, these should just never happen. If the code is invalid code, a proper error message should be generated... perhaps circular already knows which error it is though... and I vaguely remember there were some fixes in 2.7 to address incorrect code generation when optimization is on.

Well sometimes the compiler is right to point out an error, sometimes it is wrong.

In this case, I see no error at bgrabitmaptypes.pas(2702,1), so I suppose the compiler is wrong. There are some problems with "record" type.

[BigChimp]

If the bug hasn't been reported yet, it might help to report it...

I searched for 2012090607 and internal error 2012090607 in the bugtracker (including closed bugs) and didn't find anything...

[parcel]

Code: [Select]

          LOC_REGISTER,
          LOC_CREGISTER :
            begin
{$ifdef cpu64bitalu}
              if (para.size in [OS_128,OS_S128,OS_F128]) and
                 ({ in case of fpu emulation, or abi's that pass fpu values
via integer registers }
                  (vardef.typ=floatdef) or
                   is_methodpointer(vardef) or
                   is_record(vardef)) then
                begin
                  case paraloc^.loc of
                    LOC_REGISTER:
                      begin
                        if not assigned(paraloc^.next) then
                          internalerror(200410104);
                        if (target_info.endian=ENDIAN_BIG) then
                          begin
                            { paraloc^ -> high
paraloc^.next -> low }
                            unget_para(paraloc^);
                            gen_alloc_regloc(list,destloc);
                            { reg->reg, alignment is irrelevant }
                            cg.a_load_cgparaloc_anyreg(list,OS_64,paraloc^,destloc.register128.reghi,8);
                            unget_para(paraloc^.next^);
                            cg.a_load_cgparaloc_anyreg(list,OS_64,paraloc^.next^,destloc.register128.reglo,8);
                          end
                        else
                          begin
                            { paraloc^ -> low
paraloc^.next -> high }
                            unget_para(paraloc^);
                            gen_alloc_regloc(list,destloc);
                            cg.a_load_cgparaloc_anyreg(list,OS_64,paraloc^,destloc.register128.reglo,8);
                            unget_para(paraloc^.next^);
                            cg.a_load_cgparaloc_anyreg(list,OS_64,paraloc^.next^,destloc.register128.reghi,8);
                          end;
                      end;
                    LOC_REFERENCE:
                      begin
                        gen_alloc_regloc(list,destloc);
                        reference_reset_base(href,paraloc^.reference.index,paraloc^.reference.offset,para.alignment);
                        cg128.a_load128_ref_reg(list,href,destloc.register128);
                        unget_para(paraloc^);
                      end;
                    else
                      internalerror(2012090607);
                  end
                end
              else
{$else cpu64bitalu}

It may be problem in 64bit compiler.

[circular]

But here on Windows 64, it compiles.

[parcel]

But here on Windows 64, it compiles.

It compiled in 32bit windows all optimization options at fpc 23438, working fine.

[circular]

So it's a problem with svn version.

[freeman35]

Afew minutes ago I update all svn versions and rebuild and install, No error for now,
My system :
Linux XxX 3.6-trunk-amd64 #1 SMP Debian 3.6.8-1~experimental.1 x86_64 GNU/Linux
KDE 4.8.4

Lazarus At revision 39889.
FPC At revision 23451.
lazpaint-code At revision 454.

Thank you

[circular]

Cool