[fpc-devel] DWARF Error on mipsel-embedded target

[Michael Ring]

I knew for a while that dwarf output for the target is not 100% correct, 
I now found one example of a wrongly generated debuginfo, could somebody 
please help me and fix this problem in compiler?

I see the problem both with dwarf2 or dwarf4 debug info, with dwarf2 the 
error message is a little different.

Happy easter and thank you in advance,

Michael

the program:

program hello;
const
   a : array[0..3] of longWord = (
     $ffffffff,
     $ffffffff,
     $ffffffff,
     $7fffffff
   );
begin
end.

comes up with a dwarf error on gdb:

[ring at MacbookPro pic32mx795f512h]$ pic32mx-elf32-gdb hello.elf
GNU gdb (GDB) 7.6

Reading symbols from /Users/ring/devel/pic32mx795f512h/hello.elf...Dwarf 
Error: wrong version in compilation unit header (is 16964, should be 2, 
3, or 4) [in module /Users/ring/devel/pic32mx795f512h/hello.elf]

(no debugging symbols found)...done.

Dwarf2 errormessage:
Reading symbols from /Users/ring/devel/pic32mx795f512h/hello.elf...done.
invalid dwarf2 offset 2231369728


My commandline for compilation is:

DEBUG="-gw4 -godwarfsets -godwarfmethodclassprefix"
BINUTILS=pic32mx-elf32-; SUBARCH=pic32mx
CPU=pic32mx795f512h

/usr/local/lib/fpc/2.7.1/ppcrossmipsel -MObjFPC -Scghi -Ch1024 -Cs1024 
-Tembedded -Pmipsel $DEBUG -vewnhix -l -Cp$SUBARCH -Wp$CPU -XP$BINUTILS 
-FD/usr/local/bin -a hello.pas  -al -ar -at -O2 -d$CPU -B