> This was for two different occasions, just that my linebreak
> dissapeared:
> 
> package specific optimizations aren't handled unless you do it on
> commandline.
> 
> some packages are known and documented to break with certain flags, and
> if those flags are in the common area (-O3 comes to mind)  they are
> filtered out and removed.  Some packages don't even like -march= but
> will sod off and generate segfaults the next package thats built (cause
> its used in the buildprocess) .  We fix that by overriding CFLAGS for
> those specific cases.  (sandbox is a good example. it breaks using O>1)

Ahhh...so you DO care ;- )
I suppose then there may be interest in any Flag/Package combinations I find to
be not workable.  Perhaps on gentoo-dev?

> > This Larry the Cow likes insane CFLAGS optimizations!
> 
> Yes, go play with them, but what we are protecting here isn't users who
> compete by the size of their CFLAGS, but ourselves from packages that
> don't build with either the default (-O2 -march=)  or"recommended" (not
> by me) -O3 .

If I didn't want performance I would've walked right past gentoo and complained
that the install process wasn't "pretty enough" or "idiot proof".
I do like (really like) what we have here with gentoo...I'm just looking for a
_LITTLE_ more control over it.

This brings up 1 more question (you're probably not going to like)...if there's
a custom patch (possibly private - say for preferred keybindings or something)
I'd like to apply to a package before compiling...is this possible as well?


> > This is sort of what the LFS (linuxfromscratch.org) boys told
> > me...well no, their motto is FBBG (Follow Book, Book Good).
> > 
> 
> <cut>  (since you didn't tell me what measure of "stable" was, nor what
> os said system was for I'm blatantly ignoring this)

What other OS than Linux would linuxfromscratch be?  I DO admit that I did
minimum testing on this...so I probably shouldn't have said "stable", more like
"it booted".


> > I really can't think of a package that would benfit overall system
> > performance from optimization more than glibc - it's used by
> > EVERYTHING!
> 
> Ahh, so thats why theres processor specific assembler all over glibc.
> Glad to know. 

It's not really "all over" glibc.  Most (if not all) is contained in
glibc-version/sysdeps (also glibc-version/linuxthreads/sysdeps) and is included
as needed based on your HOST setting.  You may have looked further into this
than I but I'm pretty sure I didn't imagine it compiling and working for me
(though it hasn't been tested thoroughly).

--
[EMAIL PROTECTED] mailing list

Reply via email to