Fix two problems with previous commit.
authorGeorge Koehler <xkernigh@netscape.net>
Mon, 6 Feb 2017 18:19:01 +0000 (13:19 -0500)
committerGeorge Koehler <xkernigh@netscape.net>
Mon, 6 Feb 2017 18:19:01 +0000 (13:19 -0500)
commite947ce7a7fbdf554ebcde07649dadbcbdcd50690
tree19c620124115188bc9a57ebf3bbbdfde908de552
parentd4b5f6a284e11be39238ea1aabcaa9558262a4bd
Fix two problems with previous commit.

First, I can't use relonami to compare symbols, because relonami
doesn't have different values for different symbols until pass 3.  I
was wrongly pairing a hi16 or ha16 with a different symbol's lo16.
Fix by comparing item pointers.  Hack mach/proto/as/comm7.c to pass
each item pointer to mach/proto/as/mach5.c

Second, I broke hi16[...] and ha16[...] for absolute values.  Fix by
checking hr_head with DOTVAL and DOTTYP.

Having fixed the second problem, I can now switch li32 to use
eval_hl() and emit_hl(), without breaking programs that li32 an
absolute value.
mach/powerpc/as/mach0.c
mach/powerpc/as/mach1.c
mach/powerpc/as/mach4.c
mach/powerpc/as/mach5.c
mach/proto/as/comm0.h
mach/proto/as/comm7.c