[gentoo-amd64] Re: About gcc 4.1.1

[Duncan]

"Boyd Stephen Smith Jr." <bss03@××××××××××.net> posted
200608310722.30857.bss03@××××××××××.net, excerpted below, on  Thu, 31 Aug
2006 07:22:30 -0500:

> On Thursday 31 August 2006 06:43, Simon Stelling <blubb@g.o> wrote
> about 'Re: [gentoo-amd64] About gcc 4.1.1':
>> Boyd Stephen Smith Jr. wrote:
>> > Yeah, but -march might be filtered by the ebuild and -mtune not; I
>> > like using both of them.  Similar to this, I have multiple -O flags in
>> > my CFLAGS so the late non-filtered one is active.
>>
>> This is not an argument, because when filtering such stuff, we use
>> 'replace-cpu-flags' which filters all of the -march/-mtune/-mcpu
>> options. There's really no point in having them both.
>>
>> The same goes for the -O filtering.
>
> Damn those pesky developers. ;)
>
> When will they learn the solution is not to mask out CFLAGS it's to produce
> a patch so the package works with the user's desired CFLAGS. :P

Which developers you talking about, Gentoo or upstream?

If Gentoo, that's exactly what the various replace-flags and filter-flags
do, at the Gentoo level, which is the ebuild.  If you have a big desire to
avoid it, you can always copy the ebuild to your overlay and
comment-out/delete/whatever the call to the function you don't agree with.
(I did that routinely with the old monolithic xorg, for instance,
commenting out the strip-flags call, as mine aren't normally unreasonable
and when they don't work anyway for some reason, I can take appropriate
action.)

Also note that by definition, Gentoo isn't going to be removing something
as specific and basic as -march=amd64, without a very serious bug
triggering the removal.  Random -ftouchy-optimization, you have a point
(and I do the same with certain specific flags on by default based on my
-Ox setting), but removing -march=amd64 could easily screw up more than it
fixed in many cases, so they simply aren't going to do that except under
very specific circumstances.

Upstream... well, Gentoo does submit a lot of patches for such things
upstream.  Of course, whether they take Gentoo's suggestions or not
depends on the particular package and upstream devs, but Gentoo tends to
be better than most, I believe, in this regard.

Other than that, how will the upstream dev know, as long as it works for
them, until somebody submits at least a bug, and possibly a patch if the
dev isn't familiar with the arch or target platform in question?

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-amd64@g.o mailing list