[CFT] kill darwin encode_section_info

[Richard Henderson]

Gets rid of lots of Very Wrong frobbing of the interned
strings in the symbol table.

Instead we use UNSPEC patterns to record that a given type
of addressing is desired (base-relative or non_lazy_ptr)
and delay creating the backing data structures for these
until the instructions are actually emitted during final.

Since I don't have a darwin system myself, I can't test
more than just cross-compilation.  I have a little test
program that generates one of every kind of relocation,
and that seems to work, but...

Anyway, I'd appreciate someone testing this for real.


r~
-------------- next part --------------
A non-text attachment was scrubbed...
Name: d-darwin-5.gz
Type: application/x-gzip
Size: 27620 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20030512/08e49f6c/attachment.bin>