[sdcc-devel] Getting to 2.2.0 / gbdk3-2.92 / moving targets

[Michael H. <mic...@ea...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> however the fix for "&a[n]" was a little easier the lavaluereq flag
> would be incremented

On the gb, something like:

const unsigned char *film[] = {
  &door1_tiles[0x0C*0],
  &door2_tiles[0x0C*0],
  &door3_tiles[0x0C*0],
...

Only sets the first byte.  i.e it generates:

	ld	de,#film + 0
	ld	a,#<(_door1_tiles+0)
	ld	(de),a
	ld	de,#film + 2
	ld	a,#<(_door1_tiles+0)
	ld	(de),a

That was this morning - now I do not know.  Also something is trashing
memory since this morning - compiling galaxy.c in the gbdk package (theres
a tarball on sourceforge under gbdk) segfaults with a corrupted symbol.  I
dont know whos fault is it yet :)

Has anyone seen this before?  Is it truly trashed memory?  Here is gdb's
backtrace:

(gdb) backtrace
#0  addSet (list=0x20, item=0x8101550) at SDCCset.c:133
#1  0x805c973 in allocIntoSeg (sym=0x20) at SDCCmem.c:250
#2  0x805cb3e in allocGlobal (sym=0x8101550) at SDCCmem.c:363
#3  0x805d501 in allocVariables (symChain=0x8101550) at SDCCmem.c:748
#4  0x8049607 in yyparse () at SDCC.y:128

You can see that sym in (2) is currupted...

 > in this case fixed it and checked things in ...fixed a couple
of other > stuff too.
> 
> s51 is abending with memory fault when I use "-r " option so cannot
> use it from the debugger ..looking into this...
> 
> Michael anyplans for a z80 library ..or do you have them separate from
> the compiler.
The gbdk-lib package on cvs.gbdk.sourceforge.net has support for the z80
as well.  I'm not sure how to handle this yet :)  We can always hack
something together for 2.2.0 or tell them to get the libs seperatly.

> I will fix the makefiles for the z80 assembler & linker sothat they
> build with .configure ....
I'd prefer not to - amost all of the rules are in Makefile.common and are
inherieted from there.  Using a shared rules file makes the Makefiles more
maintainable & cleaner.

Blah blah.  Trying to release gbdk3-2.92 today - I just have to try and
keep up with Sandeep :)

- -- Michael

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE4wvEOUejL3SuzxEgRAlyVAKCqDuatqrCwMgkjTzExfKwAtWmpxgCfYXAt
utjogDc5imd7G6n+jnTJh/o=
=ry7K
-----END PGP SIGNATURE-----