gentoo-ppc64

beta 1 2004.3 notes

2004-10-20T22:06:00.000-05:00

Quickly

1) mac-fdisk appears to work again (yeah)

2) with us keyboard you have to use the enter key on your number pad. This is specific to the livecd, i have a bug open for it.

3) 7,3 and 7,2 powermac, rack mac should work

4) you need to grab the portage snapshot from dev.gentoo.org/~tgall to go with the 2004.3 stages. I have some "crap" to push into portage. I'll do that tomorrow.

5) umm i think that's it.

6) o yeah ... have fun

7) post bugs! bugs.gentoo.org

8) o yeah ... enjoy!

kde emerges!

2004-09-04T09:25:00.000-05:00

Greetings Programs!

Well that's probably the big news of the day. Finally got things to a point where you can emerge kde successfully! You need your ACCEPT_KEYWORDS set to ~ppc64 but hey I don't plan to leave it that way for long! We'll be stable soon.

On the livecd 2004.2 release front, I think I am going to call it good and GA. Basically this will leave PowerMac 7,3 somewhat out in the cold as well as probably the new G5 iMacs. But that's what 2004.3 can be for ;-). Besides benh hasn't put the hardware support together for these exciting pieces of new hardware. In addition I really need to put some time into getting iSeries and JS20 stuff nailed down.

Thankfully in order to focus on that, the bug load is actually back to being reasonable! I don't think I remember there only be 30-ish some bugs for a long time!

Beyond that, its laborday weekend around these parts and xorg is getting ready for a new release. Looks like it's got a few problems tho. :-/ Initial tests haven't been yielding good results, but thankfully alot of the ppc64 patches have made their way in, so it at least builds and doesn't seg fault out of the box!

3.3-hammer gorp

2004-08-15T20:51:00.000-05:00

Short one ...

If you have any of 3.3-hammer around with 3.4.1 you will have problems. 3.3-hammer had this anoyingproblem of dropping crap here and there for c++ and well ... it'll cause problems.

Lv thanks! Unfortunately this means that there might not be exactly a smooth transition from the 2004.1 to 2004.2 profiles. ... ug!

State of the 2004.2 ship

2004-08-12T22:09:00.000-05:00

Arrrrrrrrrrrr!

Avast ye lubbers!

Ok .... watching pirates of the carribean doesn't do much for my grammer says I.

So since being back from OLS, I've been trying to cut bugs down and get a few ebuilds unmasked. And now I have a new 2004.2 livecd for pmac and new stages as well.

The current beta 2004.2 livecd does have potential issues on the xserve g5 which has broadcom for networking hardware.

Tomorrow I should have a new 2004.2 livecd for pSeries boxes and for the pmac again with a newer kernel config which should be in better shape for the xServe G5. Not optimal yet but benh has a few changes coming down the pike, that will be desireable. (Updated fan control and a fix for the sungem induced freeze that can happen on xserves IF you build the sungem into the kernel)

Beyond that, should have a rs/6000 44p 270 bought here fairly soon. Going to be nice to have some hardware so I can test pSeries support directly!

bugs, BUGS, X, gnome and OLS

2004-07-23T20:14:00.000-05:00

Greets,

Well as you've probably noticed I've cleared a few more bugs while up here at OLS, least in the rare few moments that I've had in the evenings or at sessions.

I also have been working on a "stable" version of gnome. I expect to mark a 2.6 version as "stable". Tho that perhaps is a bit of a stretch, for a first intial version of gnome it doesn't seems like too much of a stretch of imagination. Course now people are going to ask for kde. :-)

To add to that, I've been a bit sick the past couple of days. Do you have any idea how much it sucks to be on the road and sick? Alot. But o well....

The X sessions here at OLS for me have been the highlight. There's going to be some serious X advancements in the next year or so and I tell ya we're long over due for it. But largely the good news is it's just simple matter of code. If anything of concern it's the status and binary X drivers and what that means to architectures that are non-x86. It might suck to be us for the purposes of gl integration into X, on which future 2d x might sit.

Well that's it for now. Later gaters!

I hate American Airlines

2004-07-20T22:22:00.000-05:00

Yes that's right. I don't all out hate too many things in this life but American Airlines just rose to the top. Here I sit in Ottawa... sans luggage. I filled out the forms. I called the #, they said, we've traced your bag, it's on the next flight, you'll have your luggage this evening. Well it's this evening, no bag.

If they hadn't told me this at approx 4pm while the stores where still open and I could have gone out and actually bought a thing or two so I'd be ready for tomorrow, but o no, let's just lie to the customer.

Least I've got my laptop, no way in friggin hell I'd let them handle that. EVER!

Beyond that Canda is nice. Looks to be a good conference ahead this week. The Ottawa Linux Symposium is always a good time.

Off to OLS

2004-07-19T21:59:00.000-05:00

Nothing like a 7am flight to get out of dodge.

So I won't get much sleep tonight... yea! Nothing like starting a long week tired but o well. Not much I can do about that.

O ... mega negitive points against Wells Fargo banks. The buggers don't sell Canadian money anymore. Gee thanks! After buying Canadian money there I don't know how many times they appearantly stopped doing that.

Well I should be uploading the latest livecds today for the 2004.2 release but genkernel has troubles in devfsd, .. thanks plasmaroo! So that pretty killed any chance of getting something rolled together and tested before taking off for the week. Guess we'll be late for the release. O WELL!

Beyond life's peachy!

livecd time (no pressure)

2004-07-18T10:28:00.000-05:00

Nothing like a joy of deadlines....

As Douglas Adams wrote, "I like the whooshing sound they make as they fly by". And it's probably going to happen that way too. I'm off to OLS quite soon and July 25th is the day when for release purposes we want all the beta livecds out for the 2004.2 release. Basically the last day I'll have access to hardware is July 19th.

So nothing like a rush job to put it all together.

The 3 stages are assembled. Built that yesterday. Have a new yaboot-static for newer pmac boxes. Just need to get the G5 XServ patches for 2.6.7 and build away. Hopefully one of the tools such as squashfs or ziofs or something actually works this time. Last time I ended up shipping a raw uncompressed fs on the cd. It work but it was a bit of a pain getting here.

Anyway... worse comes to worse, I'm digging in my heels and shall not ship,reguardless if others whine before it's time :-)

Regards,

Tom

stage and bug-o-rama

2004-07-17T12:52:00.000-05:00

Greetings,

Xorg is now marked stable. It works quite well on ati and nv hardware. Having a native version of X is quite quite handy to have.

I've been trying to get the bug list down. It was popping over 100 at time and alas there's alot of bugs that are just packages needing marking. Good timing for it given that it's nearly 2004.2 release time.

Speaking of which I'm busying building new stage and new livecds. Hopefully get all this put together soon with support for the newer g5 xserves which as I understand don't boot on the current media given we're missing a patch from benh.

Beyond that, next week is in OLS. We'll see how much I can get done from there. Going to try aweful hard to have the livecds done as i won't have any boot hardware up there to test them on.

Later!

1 week to go

2004-07-05T10:21:00.000-05:00

Hi All,

Well i'm sure you've noticed a bit of a slow down. The show I've been in just closed this weekend and the show I am directing opens this friday so it's been busy busy busy with me. As such my gentoo hacking has been very confined and focused but there has been some movement on my part.

Currently I have a new set of livecds I'm working on based on the 2004.2 profile with some additional goodies included. (lvm2, raidtools)

Beyond livecds we've got a fairly large amount of bugs open at the moment with alot of "working" software. This is a good development.

On the X front we'll have just remote X for awhile I'm afraid. We really do have to change the elfloader for ppc64 native x to work. But that's ok, this is a task that shouldn't be hard. Really it's just a copy job from the glibc elf loader into xorg's elf loader and that'll be that.

Anyway... Come one week from today my weeknights will be free again and I plan on getting alot done!

Theatre and other fun

2004-06-19T16:25:00.000-05:00

Well it's been a long week. For those that don't know I do theatre from time to time. Currently I'm in a melodrama here "Nostrums and Nonesense" and we had opening night last night to a reasonably sized crowd. This unfortunately also puts a little pressure on my hacking hours as such the bug list has gotten a bit long.

So I'll be in this show for this weekend and then 2 more weekends. But on top of that starting Monday is rehearsals for the Melodrama I am directing. I'm really going to be tied up sold for another 3 weeks effectively. Nothing like an elevated stress level for a few weeks. But hey it's fun!

And the Ottawa Linux Symposium isn't far away. (If you'll be there, hollar, also fun to meet folks and do some late night hacking )

So it would appear by this: http://anti.cute-ninjas.com/wp/index.php?p=11 that anti got native X actually working as compared to the remote display support which is currently working now. Great work. Now to figure out how to entirely and always use the dlopen stuff instead of X's elfloader which is crap.

Beyond that ... working on the bug list today, getting that shorter. Fun fun fun!

Look for livecds for 2004.2 in the next couple of weeks. Time to get that going again and this time have better support for lpar, serial console and iseries systems.

Later!

boldly pushing ahead!

2004-06-12T18:16:00.000-05:00

Well the 2004.2 profile seems to come out clean, the stages seems like they are in good shape. This is a good positive development and the earlier 3.3.3-hammer compiler problems are gone, and hey even vim is behaving!

Now it's time to generate some livecds and this time include the code for lpar boxes. (hvc support was missing prior)

Also ... it would appear the xorg code is going to have to be from their cvs snapshot. The base release appears to not build. So I'll package that up tonight in a ebuild and post that to my page on dev.gentoo.org. Ought to be good enough to at least start things off.

X marks the spot

2004-06-07T23:49:00.000-05:00

Well brad_mssw again makes a huge contribution by pointing me to this link http://freedesktop.org/bugzilla/attachment.cgi?id=183&action=view which effiectively (with a little spit and polish) resulted in a viable 64 bit X that well is looking mighty beautiful.

I have another build running on my machine otherwise I'd try out a true native X all the way so instead I am wimping out and just running a 64 bit x logo on a 32 bit ppc xserver.. but DAMN if that app running isn't the pretties thing I could have imagined.

Anyway, it's midnight I have an early morning tomorrow and have stayed up far to late. Stay tuned!

Delete me please!

2004-06-07T07:28:00.000-05:00

Greetings programs!

Well the Tonies were on last night... I have a sudden urge to travel to New York and overdose on theatre... o wait I have rehearsal tonight. buggers!

Anyway, been working on a couple of things over the weekend besides the deck, from which I have earned a nice bright red apple appearance. Nothing like the feel that you are glowing like a radioactive substance.

I've started to dive into the elf loader in xorg. This by appearances is the major portion of code that is missing for a 64 bit x. And it doesn't help that the ppc64 elf abi is just anoyingly different in some respects. So it's just not a matter of taking advantage of other implementations. Thankfully the loader in glibc also performs much the same function.

The other side of the coin has been working on various stages. Specifically stages based on 3.4 and much much newer glibc and the road for creating a livecd for IBM rs/6000 hardware. The later has been going fairly well and I suspect by about Wednesday should have something put together. The 3.4 live cds hasn't been going so well.

Q dramatic music!

It would appear there is some bug rolling around in portage or glibc. For some reason if you emerge glibc where it upgrades in a system, things are ok. But if you are using catalyst and building stages what has been happening is, baselayout would come in, create /lib and then symlink /lib64 to it. Then glibc would come along and suddenly /lib64 would turn into a directory. Bad mojo!

See /lib64 and /usr/lib64 aren't used by much of anthing out in the big wide world and as a result, some symlinks to the more sane /lib and /usr/lib and things are a-ok.

Anyway... the problem appears to be in glibc afterall. The 20040529 version to be exact. But 20040605 seems to be a-ok. (Least thus far :-/)

More as it happens!

To 3.4 or not to 3.4...

2004-06-03T06:52:00.000-05:00

Here we are early on in the life of gentoo-ppc64 and have a quandry already. From initial tests, 3.4 seems to be ok. Tho on the other hand from the general maintainers point of view 3.3-hammer is more interesting in the short term. Certainly for 2004.2 we will do 3.4 as is the entire gentoo release but what to do with the 2004.1? Leave it or move it up to 3.4 as well?

Beyond that... been busy marking packages and going after the bug list, trying to keep that down. It's a constant battle!

O general warning, next week and going on for about 6 weeks I'll be kinda distracted. Have play practise in the evenings. Go figure, I do theatre :-). If you're in Mantorville, come see. :-)

Productive Day...

2004-05-31T23:08:00.000-05:00

Well what a day....

News hits slashdot.

Wife and kid go the cities.

I get the pseries iso working via helpful folks that have p615 and p630 hardware. (Some fellow on a p610 is having a problem but that looks like more of a firmware thing)

I get the garage cleaned. (yeah!)

And I fixed a number of bugs. The coreutils one looks to be the most important one of the day. uname -a was core dumping and that was causing havoc as you can imagine.

I dived into the xorg code and got it to build ... but debugging it lead me quickly to the elfloader of theirs. I suspect there is some code there that needs to be written as they have some specific sections by architecture... and as you can guess there's no __powerpc64__ bits! Fun!

Time for bed.... ZZZZZZZZzzzzzzzzz

fame?

2004-05-30T12:21:00.000-05:00

Well not really. No way ... ain't gunna happen, but it's fun to see that http://osnews.com /story.php?news_id=7204 has the beta announce up. I half wonder if this'll make /. . Umm that might be scary. Anyway, hopefully the one good thing that'll come of the news is more users and more testers which is always a good thing (tm).

So what's news? Well the pSeries livecd is currently building and then after that I'll probably do some optimized stages for power3. (Well maybe)

Then once pSeries is done I'd like to get iSeries added in real quick and then it's off to work on 2004.2! Which from a toolchain perspective seems to be somewhat promising. gcc 3.4, glibc 20040529, and binutils 2.15 appear to be viable from reports.

Well gotta run. Later!

Success!

2004-05-28T22:29:00.000-05:00

Greetings Programs!

Well it's been a long road the past few days and I've been having all sorts of fun fixing and running into various bugs. But it's been well worth it, and now it's starting to pay off.

There are new stages out on dev.gentoo.org/~tgall. I've also placed a new g5 livecd out as well. This cd works! The version of mac-fdisk on the livecd however is b0rked. So you'll ahve to partition your disk via other routes. (Via mac osx, or via the ppc32 gentoo-livecd.) I am as I type this building a new livecd for the G5 with a newer mac-fdisk that should be better. Least in initial testing it did at least display the partition table correctly. But that's not to say it's perfect. BIG HUGE thanks to tchiwam for putting the time into mac-fdisk.

I'll have a new pseries/rs6000 livecd by Monday at the latest, Saturday at the earliest. I've been trying to buy a pseries 270 off of ebay but I can't believe how expen$ive they are! It would be nice to have for testing but the suckers are going for a couple of grand!!!! This for a dual Power3 375 Mhz 630+ box. OUCH!

Next up in the Q is iseries support, chrp support (for ppc32) and xorg x11 support. I'm quite excited about the prospect of X11 support because this is really the turning point for me where gentoo-ppc64 can self host it's own development.

Enough for one night! Take care!

libperl, new stages and other fun

2004-05-19T21:53:00.000-05:00

Well it's been a longer couple of days. First with the fixes to ameslab I have a new viable 2.6.6 kernel which is a good development. (Wanna see what patches are going into ameslab? Point your webbrowser to http://bk.linuxppc64.org)

The one fly in the oitment (and there is always one with this project), is that this also requires the 20040408 or greater version of the 3.3-hammer gcc compiler. Joy... nothing like reving the toolchain so you have to build all the stages over. So I started down that path and the next boogie man popped up.

libperl. So it turns out it doesn't want to build squaking about main not being defined while it was building the .so.. which is entirely bogus since first there shouldn't be a main and second it was missing the shared flag. wtf!

Emerging the package from the command line and it's fine. Yet oddly on another machine in another chroot building the same stage works just fine. So I'm wondering if something isn't h0rked with my partition I use for building. Dunno ...

At the same time, 3.4 is also very interesting so it's time to get that rolling, not to mention also start to get more useful user packages built.

So much to do!

O I love C

2004-05-15T10:17:00.000-05:00

Greetings Programs!

Well last night I was able to finally get an iso put together and get it burnt. (Hint when using imation CD-Rs 700 Meg Media, burn at like 2x or you'll end up with a coaster)

Booting up the 2.6.6 kernel on my g5 test subject however didn't fair so well. The kernel died. O Joy. Rapture.

Last message is smp_core99_kick_cpu done

That's fairly early in the boot process.

This does seem to be an ongoing issue in that there is no clear direction as far as when / if a reasonable piece of code is reasonable, be it a kernel, compiler or what have you at least in the current ppc64 ecosystem. There *USED* to be such a setup. (I was the first ppc64 kernel maintainer) I had a stable and a devel and would basically let devel swing back and forth between liberal and conservative patch acceptance based on if I was about to upgrade stable to the next stable release. This worked. People liked it.

Anyway so closer to the problem at hand. We have a couple of variables here.

1) 2.6.6 itself could be broke
2) gcc 3.3-hammer might be the issue

gcc for the current gentoo-ppc64 is based on a 20040215 cvs snapshot. It's been a reasonable compiler.

The gcc I used prior was from back in december, that also was a reasonable compiler.

In either case a working kernel did not result from using those compilers.

So... I built the gcc-20040408-r1 ebuild which is a snapshot from 3.3-hammer and from 20040408 (thus the name). And then hit this little gem in the ucl code WHICH we need since that's for gcloop.

In file included from /usr/include/features.h:295,
from /usr/include/limits.h:26,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/limits.h:122,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/syslimits.h:7,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/limits.h:11,
from include/ucl/uclconf.h:46,
from include/ucl/ucl.h:39,
from crypto/ucl_compress.c:22:
/usr/include/sys/cdefs.h:183:1: warning: "__attribute_pure__" redefined
In file included from include/linux/compiler.h:20,
from include/linux/init.h:5,
from crypto/ucl_compress.c:19:
include/linux/compiler-gcc3.h:22:1: warning: this is the location of the previous definition
In file included from /usr/include/features.h:295,
from /usr/include/limits.h:26,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/limits.h:122,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/syslimits.h:7,
from /usr/lib/gcc-lib/powerpc64-unknown-linux-gnu/3.3.3/include/limits.h:11,
from include/ucl/uclconf.h:46,
from include/ucl/ucl.h:39,
from crypto/ucl_compress.c:22:
/usr/include/sys/cdefs.h:192:1: warning: "__attribute_used__" redefined
In file included from include/linux/compiler.h:20,
from include/linux/init.h:5,
from crypto/ucl_compress.c:19:
include/linux/compiler-gcc3.h:17:1: warning: this is the location of the previous definition

Soooo something changed in the newer compiler as far as the includes are concerned and well ... there you go.

So that's the compiler side.

Now 2.6.6 itself from ames could be to blame. That's right possible. Tho there hasn't been any commits to the tree (you can find out by visiting http://bk.linuxppc64.org) lately so I would expect that if something were broke, generally we'd see some sort of Changeset to address it.

So that's specifically the problem.

But this does underscore a larger problem and lament that i have with the ppc64 ecosystem. Community based distros such as gentoo or debian are basically going to have a tough go of it in the current atmosphere. This does not bode well. I think we need to create a feedback loop between "stable" and code under "development" so that folks that want to use a community distro can, those that put together a community distro can, and those working on the code have a better feel for when their code is or is not working so well.

It's hard to see what the commercial distro types might also experience in this area if it's a challenge for the community distros. Anyway, hopefully this is one of those elements that will improve in time.

blk_run_queues

2004-05-13T22:17:00.000-05:00

So the adventure this evening was centered around pulling gcloop support into the ppc64 ameslab kernel sources. For livecds of course this is the preferred feature for mounting the file system off of cd.

This went pretty smooth as far as the gcloop patch is concerned BUT, there's a reference to blk_run_queues which was removed from 2.6.x. LWN story on unplug change

So there's a bit of work to do there, and hopefully it'll be well worth it.

2.6.6 ppc64-sources

2004-05-13T12:53:00.000-05:00

So I created a new ebuild for 2.6.6 from ames for ppc64-sources. There were a couple of old anoying things tho that had to be dealt with and it seemed that a jump up to 2.6.6 was due.

1) the fool 3.3-hammer gcc compiler from 0215 seems to not like to find stdarg.h when --nostdinc is set.

2) I had CROSS-COMPILE set in the 2.6.4 ppc64 kernel sources

3) 2.6.4 is crusty

So the new ebuild is making it's way to portage. The 2.6.6 ames patch is making it's way to the mirrors and catalyst seems a bit happier as it builds new kernels with genkernel. Now I'm slighty worried about the gcloop stuff and the cds that'll result as I need to create a tarball with the top level setup on the cd. But details details. We're getting there!

On the subject of Catalyst

2004-05-12T07:54:00.000-05:00

So I spent the majority of last night of trying to figure out what the &*()^%&%*^ catalyst is doing when it silently bombs making stage 2 livecd images. From the information it dumps to the screen it emerges the ppc64-sources like it should, prints G5 to the screen and then decides it's had a long hard day and calls it quits without so much as a reason.

Well here's what was wrong and this subtile bit of trivia was about as much fun as a game of logic with a 2 year old. It turned out that in the ppc64-sources appearantly in some spout of stupidity (well I knew this problem was my fault anyway!) I had added -ames- to the kernel version name to accurately indicate it was a pull from the ameslab kernel tree.

So in an ideal world /usr/src/linux should be a sym link to the unpacked kernel source tree and well ... it would have been had I not added ames back then... DUH! And that my friends is why catalyst was silently calling it quits.

Beyond that I threw in the dirdiff package into dev-utils. Great tool, highly recommended!

So I'm back on the road of progress with the livecd creation process, I am going to upgrade the kernel version, and we'll go from there!

So it begins...

2004-05-11T18:46:00.000-05:00

Greetings,

I've decided to start to blog my gentoo-ppc64 activities. As the gentoo-ppc64 lead I figure this is probably of some value so folks can see what I'm up to and what plan I have (if any). This will sort of track on a day to day basis what's up.

So with that #define ENGAGE_FISH_BOWL