"Nadav Har'El" <[EMAIL PROTECTED]> writes:
> But if you're going to compile some of the code with -O1 and -O3, your
> Makefile or configure script are going to look realy hairy :( It will
> look like black magic. A better (but more time consuming) thing to do is
> to try to find the offending piece of code (which compiles wrong) and send
> a bug report to gcc and/or try to circumvent the bug.
I agree completely, but just for the hell of it, I don't think there
will be much black magic in the Makefile. Here is one possible make
hackery to skin this particular cat (UNTESTED, GNU make):
OBJ_O1 := foo.o1 bar.o1
CFLAGS_O1 := $(CFLAGS) -O1
OBJ_O3 := xyzzy.o3
CFLAGS_O3 := $(CFLAGS) -O3
target: $(OBJ_O1) $(OBJ_O3)
$(CC) -o $@ $^ $(REST_OF_LINK_CMD)
%.o1: %.c
$(CC) -c $(CFLAGS_O1) $(CPPFLAGS) $< -o $@
%.o3: %.c
$(CC) -c $(CFLAGS_O3) $(CPPFLAGS) $< -o $@
As mentioned before, it should be perfectly fine to compile various
files with different optimizations (or without optimization at all)
and then link them together. -O3 is especially (mostly) harmless,
since the only thing it does compared to -O2 is inlining. Can that be
affecting some of your code, btw?
--
Oleg Goldshmidt | [EMAIL PROTECTED]
=================================================
"We work by wit, and not by witchcraft,
And wit depends on dilatory time."
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
