[Solved]Cannot build x86_64 compiler with fpcup.

[parcel]

Code: [Select]

C:/development/fpcbootstrap/rm.exe -f ppcrossx64.exe
C:/development/fpcbootstrap/gmkdir.exe -p x86_64/units/i386-win32
C:/development/fpc/compiler/ppc.exe -Ur -Xs -O2 -n -Fux86_64 -Fusystems -FuC:/de
velopment/fpc/rtl/units/i386-win32 -Fix86_64 -FE. -FUx86_64/units/i386-win32 -dR
ELEASE    -dx86_64 -dGDB -dBROWSERLOG -Fux86 -Sew pp.pas
pinline.pas(361,38) Error: Ordinal expression expected
pinline.pas(363,42) Error: Ordinal expression expected
pinline.pas(632) Fatal: There were 2 errors compiling module, stopping
Fatal: Compilation aborted
make.exe[3]: *** [ppcrossx64.exe] Error 1
make.exe[3]: Leaving directory `C:/development/fpc/compiler'
make.exe[2]: *** [cycle] Error 2
make.exe[2]: Leaving directory `C:/development/fpc/compiler'
make.exe[1]: *** [compiler_cycle] Error 2
make.exe[1]: Leaving directory `C:/development/fpc'
C:\development\fpcbootstrap\make.exe: *** [build-stamp.x86_64-win64] Error 2
C:\development\fpcbootstrap\make.exe: Leaving directory `C:/development/fpc'
Info: 2013-04-15 10:30:13: fpcup finished.
An unhandled exception occurred at $004057C1:
EAccessViolation: Access violation

After FPC trunk r24241, fpcup cannot build x86_64 cross compiler.

There is a two error, it may caused by r24241.

http://svn.freepascal.org/cgi-bin/viewvc.cgi?view=revision&revision=24241

Is it compiler problem?

[BigChimp]

Looks like a problem in the source (as opposed to fpcup), yes.

To make sure you're not dealing with stale ppus (a problem with fpmake), you can delete all ppus, .a, and .o files etc as indicated by the help message.

[parcel]

I tried again, failed.

Code: [Select]

360 found:=false;
361 for i:=0 to high(variantdesc^.branches) do <--- Error
362 begin
363 for j:=0 to high(variantdesc^.branches[i].values) do <--- Error
364 if variantdesc^.branches[i].values[j]=tordconstnode(p2).value then
365 begin
366 found:=true;
367 variantselectsymbol:=tfieldvarsym(variantdesc^.variantselector);

It maybe compiler didn't recognize

Code: [Select]

variantdesc^.branches and

Code: [Select]

variantdesc^.branches[i].values as array types or something wrong. 

[martinrame]

I can confirm this can be reproduced with bootstrapping versions 2.6.0 and 2.6.2.

Does anyone filled this issue in Mantins?.

[BigChimp]

Note: if somebody enters this issue as a bug in mantis, please make sure you can replicate the problem using regular make distclean all etc as fpcup is not part of fpc and is not supported by them.

That said, lately there was a bug in 64 bit windows fpc that was discovered by using fpcup... and solved by the fpc/Laz guys.

[parcel]

It is fixed at 24260. 

[BigChimp]

Great - thanks for posting that.