danger: current GCC makes a hash out of initfini.c

[cgd@broadcom.com]

I was trying to compile glibc srcs as of about 2003-10-02 with GCC
sources as of a few days earlier, and it looks like current GCC is
totally shredding initfini.c.

It seems to be reordering a lot of the contents of the file (e.g.,
ordering of functions relative to other stuff, it looks like).

-fno-reorder-functions -fno-reorder-blocks didn't seem to help, but
-O1 (instead of -O2) at least made things *look* normal.

(this is for mips64-linux.)


It looks to me like this code is *really* quite fragile (to put it
nicely 8-).  Is there other code like this that's likely to break as
well?  (I suppose I might find some of it when I restart my build...)


Have others seen problems like this?


chris