1 |
John Myers posted <200602081850.05729.electronerd@×××××××××××××.net>, |
2 |
excerpted below, on Wed, 08 Feb 2006 18:49:54 -0800: |
3 |
|
4 |
> On Wednesday 08 February 2006 16:41, Kevin F. Quinn (Gentoo) wrote: |
5 |
>> On x86 this is not enabled since it interferes with debugging (info gcc |
6 |
>> isn't too clear on this, it's mentioned on the description of '-O') - |
7 |
>> is this different from amd64 (I'm still x86-only :/ )? |
8 |
|
9 |
> x86_64 does not require frame pointers for debugging, so |
10 |
> -fomit-frame-pointer is enabled with -O |
11 |
|
12 |
/Click/. THAT'S the reason I hadn't removed the potentially redundant |
13 |
-fomit-frame-pointer! I'd forgotten about that, but the manpage /does/ |
14 |
mention that it's only enabled with -O<whatever> where it won't interfere |
15 |
with debugging, and I was unsure of whether that was the case on |
16 |
x86_64/amd64 or not, so I left it in. Knowing that it doesn't interfere |
17 |
with debugging on amd64, and is therefore enabled by -O<whatever>, I could |
18 |
probably remove the redundancy, now. |
19 |
|
20 |
... Or rather, I could, if I was running a no-multilib subprofile. I |
21 |
don't do a lot of x86 compiling (and have a policy of not running |
22 |
binary-only stuff either, so that means I don't have much x86 32-bit on |
23 |
my system), but as I'm on the standard multilib-by-default profile |
24 |
(actually, I'm already on the still unreleased 2006.0, having diffed it |
25 |
and 2005.1 and decided I might as well do so when I finally upgraded from |
26 |
2005.0, skipping 2005.1 entirely), gcc/glibc/sandbox and the like have |
27 |
32-bit compiled elements, as does grub, of course, so maybe I better leave |
28 |
it in, after all. |
29 |
|
30 |
Actually, grub is really the only thing really keeping me from going |
31 |
no-multilib profile, but as I said, I don't like running binary-only |
32 |
stuff, and that extends to grub-static as well since there's an |
33 |
alternative, which means I'm supporting a LOT of 32-bit overhead in terms |
34 |
of maintaining that multilib toolchain, all so I can compile ONE single |
35 |
program, grub, from source. OTOH, being a bootloader, grub /is/ rather |
36 |
critical to my machine's operation, even if only for a few seconds every |
37 |
few weeks or whenever it is that I decide to reboot. |
38 |
|
39 |
However, now, having had this discussion, I'll be clearer on all the |
40 |
implications of -fomit-frame-pointer, and remember exactly why I chose to |
41 |
keep it in my CFLAGS -- it boils down to 32-bit compatibility. |
42 |
|
43 |
-- |
44 |
Duncan - List replies preferred. No HTML msgs. |
45 |
"Every nonfree program has a lord, a master -- |
46 |
and if you use the program, he is your master." Richard Stallman in |
47 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
48 |
|
49 |
|
50 |
-- |
51 |
gentoo-amd64@g.o mailing list |