Mailing List Archive

On Wed, 2004-09-15 at 04:21, Paul de Vrieze wrote:
> On Tuesday 14 September 2004 00:00, Mike Frysinger wrote:
> > On Monday 13 September 2004 03:20 pm, Paul de Vrieze wrote:
> > > A popular cause
> > > is opening certain webpages in galeon (gtk based) while in konqueror
> > > it is not a problem.
> >
> > yeah i use pan to test it ... just open it up and click on some group
> > names
> >
> > also found clicking e-mails in kmail triggers it too
> > -mike
>
> It seems that adding -fno-sse2 to the compile flags fixed it.


> This would
> imply that it is definately a compiler bug.

correct.

From the reports we are getting in/about 3.4.x and SSE2 and or
fundamental CXX bugs the chances are looking pretty slim to none that
3.4.x will be targeted for the next release cycle for many/most arches.
With development being stopped on the 3.4.x series and the 3.5 being
sorta a mess I can't see us leaving the existing toolchain setup for a
while expect maybe for a push of binutils-2.15 to stable on a few
arches.

>
> Paul
--
Ned Ludd <solar@gentoo.org>
Gentoo (hardened,security,infrastructure,embedded,toolchain) Developer

> correct.
>
> From the reports we are getting in/about 3.4.x and SSE2 and or
> fundamental CXX bugs the chances are looking pretty slim to none that
> 3.4.x will be targeted for the next release cycle for many/most arches.
> With development being stopped on the 3.4.x series and the 3.5 being
> sorta a mess I can't see us leaving the existing toolchain setup for a
> while expect maybe for a push of binutils-2.15 to stable on a few
> arches.

As far as I can tell, the SSE2 bugs have only hit 3.4.2, which is now
masked. 3.4.1 appears to be fine, and indeed less broken for SSE2 than
3.3.x. The CXX bugs will only come out of the woodwork with testing, so
some testing (~x86) would be nice.

3.5 is now 4.0, and as far as I know I'm the only person doing serious
testing (testsuites included). From what I've seen so far, it's fairly
rock-solid for where it is at the moment (early stage3), certainly more
so than 3.4 was; there are still internal compiler errors on important
stuff like glibc but these invariably get fixed in the last week-long
"big push" where they desperately call for testers in a mad rush to get
the thing released! That said, there's no way gcc-4 will be in ~x86 for
2004.3, almost no way (unless it surprises me) it'll be in ~x86 for
2004.0, but may be in most ~arches for 2004.1, assuming the slightly
stronger code strictness doesn't cause to many nasty bugs and assuming
we can iron out all the ICEs.

2004.0/2004.1? time warp? 2005.0/.1?

assuming the slightly stronger code strictness doesn't cause to many
nasty bugs and assuming we can iron out all the ICEs.

^^ yeah sure like we will have more devs help test a new toolchain, you
forget it doesnt come with translucency and shadows, something that
guarantees you plenty of (dev) testers, what's gcc, just some thing
portage calls to build stuff, who would care about that?
so much easier to complain about things than help fix them


>
> 3.5 is now 4.0, and as far as I know I'm the only person doing serious
> testing (testsuites included). From what I've seen so far, it's fairly
> rock-solid for where it is at the moment (early stage3), certainly
> more so than 3.4 was; there are still internal compiler errors on
> important stuff like glibc but these invariably get fixed in the last
> week-long "big push" where they desperately call for testers in a mad
> rush to get the thing released! That said, there's no way gcc-4 will
> be in ~x86 for 2004.3, almost no way (unless it surprises me) it'll be
> in ~x86 for 2004.0, but may be in most ~arches for 2004.1, assuming
> the slightly stronger code strictness doesn't cause to many nasty bugs
> and assuming we can iron out all the ICEs.


--
gentoo-dev@gentoo.org mailing list

> 2004.0/2004.1? time warp? 2005.0/.1?

Heh, whoops, stupid Rob. 2005.0.

> ^^ yeah sure like we will have more devs help test a new toolchain, you
> forget it doesnt come with translucency and shadows, something that
> guarantees you plenty of (dev) testers, what's gcc, just some thing
> portage calls to build stuff, who would care about that?
> so much easier to complain about things than help fix them

True. But if it happens, and gcc-4.0 is in 2005.0, and all sorts of
stuff breaks because the devs ignored our pleas to test, then they can't
say we didn't warn them. Just like, well, what's happening with 3.4 and
~x86...

Robert Moss wrote:
> As far as I can tell, the SSE2 bugs have only hit 3.4.2, which is now
> masked. 3.4.1 appears to be fine, and indeed less broken for SSE2 than
> 3.3.x. The CXX bugs will only come out of the woodwork with testing, so
> some testing (~x86) would be nice.

*cough*bullshit*cough*

most of the sse2 bugs have been there since 3.4.0 and the release
manager even commented on them not being fixed (after once again
targeting them for the next release). Scott Ladd is going to be working
on them, so hopefully we'll see some fixes:
http://gcc.gnu.org/ml/gcc/2004-09/msg00936.html

The CXX bugs you're referring to only happen for nptl users, who have a
broken pthread.h which isnt parsable by g++ in any way. this is fixed in
the latest glibc snapshot, but that snapshot has a ton of issues of it's
own... like -requiring- 2.6 kernel headers, semi-broken libresolv, etc.
I'll break out a patch and apply it to older snapshots, but the result
of that snapshot was so discouraging I havent really wanted to touch glibc.


> 3.5 is now 4.0, and as far as I know I'm the only person doing serious
> testing (testsuites included). From what I've seen so far, it's fairly
> rock-solid for where it is at the moment (early stage3), certainly more
> so than 3.4 was; there are still internal compiler errors on important
> stuff like glibc but these invariably get fixed in the last week-long
> "big push" where they desperately call for testers in a mad rush to get
> the thing released! That said, there's no way gcc-4 will be in ~x86 for
> 2004.3, almost no way (unless it surprises me) it'll be in ~x86 for
> 2004.0, but may be in most ~arches for 2004.1, assuming the slightly
> stronger code strictness doesn't cause to many nasty bugs and assuming
> we can iron out all the ICEs.

O_O
http://gcc.gnu.org/ml/gcc-testresults/2004-09/msg00771.html
look at those testsuite results for amd64 for today...

though, at the rate things are going, we might have 4.0 stable before 3.4.


Travis Tilley <lv@gentoo.org>

--
gentoo-dev@gentoo.org mailing list

On Sat, 2004-09-18 at 08:02, Daniel Goller wrote:
> ^^ yeah sure like we will have more devs help test a new toolchain, you
> forget it doesnt come with translucency and shadows, something that
> guarantees you plenty of (dev) testers, what's gcc, just some thing
> portage calls to build stuff, who would care about that?

C'mon, we run Gentoo. Just tell people it optimizes their code better
and everything will run faster. They'll swarm to it.
--
Donnie Berkholz
Gentoo Linux

On Sat, 2004-09-18 at 23:20, Donnie Berkholz wrote:
> On Sat, 2004-09-18 at 08:02, Daniel Goller wrote:
> > ^^ yeah sure like we will have more devs help test a new toolchain, you
> > forget it doesnt come with translucency and shadows, something that
> > guarantees you plenty of (dev) testers, what's gcc, just some thing
> > portage calls to build stuff, who would care about that?
>
> C'mon, we run Gentoo. Just tell people it optimizes their code better
> and everything will run faster. They'll swarm to it.

Just in case anybody missed this one. (Linux C and C++ Compilers)
http://www.coyotegulch.com/reviews/linux_compilers
Gentoo linux was used for his testing.

--
Ned Ludd <solar@gentoo.org>
Gentoo (hardened,security,infrastructure,embedded,toolchain) Developer

On Sunday 19 September 2004 04:27, Ned Ludd wrote:
> Just in case anybody missed this one. (Linux C and C++ Compilers)
> http://www.coyotegulch.com/reviews/linux_compilers
> Gentoo linux was used for his testing.

The author of that article also reported a binary incompatibility between
Gentoo's GCC and Intel's icc compiler suites ... he reports that the problem
doesn't exist with "stock" GCC.

Best regards,
Stu
--
Stuart Herbert stuart@gentoo.org
Gentoo Developer http://www.gentoo.org/
http://stu.gnqs.org/diary/

GnuPG key id# F9AFC57C available from http://pgp.mit.edu
Key fingerprint = 31FB 50D4 1F88 E227 F319 C549 0C2F 80BA F9AF C57C
--

On Sunday 19 September 2004 05:20, Donnie Berkholz wrote:
> On Sat, 2004-09-18 at 08:02, Daniel Goller wrote:
> > ^^ yeah sure like we will have more devs help test a new toolchain, you
> > forget it doesnt come with translucency and shadows, something that
> > guarantees you plenty of (dev) testers, what's gcc, just some thing
> > portage calls to build stuff, who would care about that?
>
> C'mon, we run Gentoo. Just tell people it optimizes their code better
> and everything will run faster. They'll swarm to it.

Besides, while the new xorg offers all that eyecandy, it also means that x is
a lot slower than it needs to be so it is not actually practical to use it on
day to day basis. (Luckilly the slowness can be stopped very fast with
killall xcompmgr ;-)

Paul

--
Paul de Vrieze
Gentoo Developer
Mail: pauldv@gentoo.org
Homepage: http://www.devrieze.net