gmane.linux.gentoo.bsd

Re: New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Alexis Ballier — 2012-05-09T12:18:08

On Wed, 9 May 2012 01:22:50 -0400
Richard Yao <ryao< at >cs.stonybrook.edu> wrote:


probably because it gets built and installed before python2 and python
ebuilds made it the default back then...

Re: New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Richard Yao — 2012-05-09T05:22:50

Why is python3 the default interpreter if people might run into problems
because of it?

Re: New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Alexis Ballier — 2012-05-08T12:49:26

Some updates:
- python3 is set as the default interpreter, remember to switch it to
python2 or you might run into problems; anybody knows how did they fix
this for gentoo linux when python3 reached stable ?
- creating pie executables is broken: they'll get textrels on x86 and
  will simply not link on amd64; will send a patch to gcc fixing this.

A.

RES: New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Eduardo Schoedler — 2012-05-05T19:08:35

Great work, thanks!

--
Eduardo Schoedler


-----Mensagem original-----
De: Alexis Ballier [mailto:aballier< at >gentoo.org] 
Enviada em: sábado, 5 de maio de 2012 14:37
Para: gentoo-bsd< at >lists.gentoo.org; bsd< at >gentoo.org
Assunto: [gentoo-bsd] New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Hi,

I've (finally) made new fbsd stages. They are available at:
http://dev.gentoo.org/~aballier/fbsd9.0/
in their respective directories.
(at the time of writing this email, there's only amd64 but x86 will
appear soon)

I'll make some more testing and would appreciate any feedback. If
everything goes well I'll push them to the mirrors, in the usual
experimental directory.

I would like to thank especially Yuta SATOH and Richard Yao for their
work on making new stages for Gentoo/FreeBSD which made the task much
easier.

Regards,

Alexis.

New Gentoo/FreeBSD 9.0 stages for x86 and amd64

Alexis Ballier — 2012-05-05T17:36:47

Hi,

I've (finally) made new fbsd stages. They are available at:
http://dev.gentoo.org/~aballier/fbsd9.0/
in their respective directories.
(at the time of writing this email, there's only amd64 but x86 will
appear soon)

I'll make some more testing and would appreciate any feedback. If
everything goes well I'll push them to the mirrors, in the usual
experimental directory.

I would like to thank especially Yuta SATOH and Richard Yao for their
work on making new stages for Gentoo/FreeBSD which made the task much
easier.

Regards,

Alexis.

Re: New x86-fbsd stage3

Richard Yao — 2012-04-27T18:06:30

Thanks for sharing this procedure. I am happy to hear that you did a
successful installation. :)

Re: New x86-fbsd stage3

Michael Orlitzky — 2012-04-27T17:46:55

I was finally able to get up and running, although in a roundabout manner.

1. Boot the PC-BSD live DVD, and install it to the hard drive.

2. Reboot to the GhostBSD liveCD (PC-BSD panics).

3. Mount the drive, and delete everything except /boot.

4. Untar this stage3 and the latest portage.

5. Reboot.

6. Something goes wrong, /dev/ad0s1a is read-only?

7. Boot to GhostBSD, and fsck a few times. Tons of errors fixed.
   Whatever =)

8. Reboot again, it works.

New x86-fbsd stage3

Richard Yao — 2012-03-21T12:42:40

Dear Everyone,

I made an unofficial x86-fbsd stage3. It is available at the following
address:

http://www.cs.stonybrook.edu/~ryao/stage3-i686-freebsd-9.0.tar.bz2

Yours truly,
Richard Yao

Re: Gentoo/FBSD Install

Javier Villavicencio — 2011-04-12T01:07:15

First, familiarize yourself with installing vanilla FreeBSD. 
Specifically the partitioning (bsd labels and such) and boot process.

Second, boot off any FreeBSD-8.x kernel bootable media, and get a shell 
interface.

Third, unpack the stage3 and continue to install as if it was a Gentoo 
install, bear in mind the obvious FreeBSD differences (kernel 
compilation, modules loaded in the bootloader, etc).

Troubleshooting should be easy if you got the second step right.

As of now, a bunch of < at >system currently keyworded ~x86-fbsd doesn't 
compile/work as it should. Help (as in: patches, bugs with proper 
patches reported upstream) is always welcome.

Gentoo/FBSD Install

Michael Orlitzky — 2011-04-11T18:32:27

I'd like to set up a Gentoo/FBSD install to test some packages, but the
docs are out of date and I can't figure it out myself. Has anyone set
one up recently? If so, can it be explained relatively painlessly?

Re: G/FreeBSD: 8.0, (almost) new stages and removal of old versions

Javier Villavicencio — 2010-01-11T18:20:03

I've got some stuff for 8.0 done, but it's getting quite hard to work
with as the stages progress.
For example, after freebsd-lib gets compiled, portage (actually python)
stops working (it keeps emerging things until it fails then resuming
doesn't work). The reason? The new tty code on fbsd, where portage
attempts to get a tty, it's there that it dies.
I've been patching portage (2 lines) for this, but it got very annoying
at some point, plus summer and no AC keeps me away from going back to
stage building unless it's cloudy outside.
Other than that, it seems it's going to be easier to build the stages
from 8.0 instead of uprading from 7.2, which doesn't sound too good :+(

I'll check if my access to our git repo still works and upload what I
have in ebuilds there (bunch of updated patches that could be useful).

Salu2,
Javier.

G/FreeBSD: 8.0, (almost) new stages and removal of old versions

Alexis Ballier — 2010-01-11T15:00:17

Hi all,

I am going to remove all the freebsd ebuilds that are not 7.2; the 6.2
profiles have been gone for a while and ebuilds are really not
maintained, 7.1 profiles have been deprecated for a while too (2
months). If anyone has objections, speak up now!

For those of you who have not updated yet, it is probably simpler to
start a fresh install from the 7.2 stages that I've pushed some months
ago (see experimental/x86/freebsd/7.2/stages/ in the mirrors). I have
not received rants against those stages so I'll assume they're fine :)
About sparc-fbsd, I have (thanks to Monkeh) access to a sparc-fbsd box
where I've put a 7.2 but the port is more surviving than anything else.
The box doesn't have enough disk space to build stages for example,
thus I am nowhere near to be able to push new stages.

As for 8.0, I'll start playing with it now; if any of you have started
to play with it and wanna share the experience, please poke me.

Regards,


Alexis.

New profile.bashrc.

Javier Villavicencio — 2009-12-21T01:59:05

-----BEGIN PGP SIGNED MESSAGE-----
Hash: RIPEMD160

Hello everybody.

I'm back, sort of, job's workload went a bit down thanks to a change
freeze during holidays.

So I went on a emerge rampage that got me bitching and moaning about the
parallel make install problems. Here's the profile.bashrc that I came up
with (charset.alias stuff taken from the Gentoo/Prefix profile.bashrc,
I'm in doubt here, was this fixed already?).
Please review/test/etc. Right now my G/FBSD install survives an emerge
- -e < at >system with MAKEOPTS=-j6 which it didn't survive before...
Except for sys-apps/ed, which is using a bare /usr/bin/install from it's
Makefile.

By the way, anybody working on 8.0 ebuilds? ^_^
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.12 (MingW32)

iEYEAREDAAYFAksu1mkACgkQgvV6MZSadQoZ+gCdHxpAvyVCDUbxAo2Up6CsUyDb
TP8An2k5gBlIRQnmBri9+LdQZ4S4/Jdp
=YVws
-----END PGP SIGNATURE-----
if [[ ${EBUILD_PHASE} == compile ]] ; then
if grep -q "Assume that mode_t is passed compatibly" ${S} -r --include openat.c; then
eerror "The source code contains a faulty openat.c unit from gnulib."
eerror "Please report this on Gentoo Bugzilla in Gentoo/Alt product for component FreeBSD."
eerror "http://bugs.gentoo.org/enter_bug.cgi?product=Gentoo%2FAlt&component=FreeBSD&op_sys=FreeBSD"
die "Broken openat.c gnulib unit."
fi
        if grep -q "test .*==" "${S}" -r --include configure; then
                eerror "Found a non POSIX test construction in a configure script"
                eerror "The configure checks of this package may not function properly"
                eerror "Please report this on Gentoo Bugzilla in Gentoo/Alt product for component FreeBSD."
                eerror "http://bugs.gentoo.org/enter_bug.cgi?product=Gentoo%2FAlt&component=FreeBSD&op_sys=FreeBSD"
        fi
fi

# Stolen from prefix

# Hack to avoid every package that uses libiconv/gettext
# install a charset.alias that will collide with libiconv's one
# See bugs 169678, 195148 and 256129.
# Also the discussion on
# http://archives.gentoo.org/gentoo-dev/msg_8cb1805411f37b4eb168a3e680e531f3.xml
bsd-post_src_install() {
        local f
        if [[ ${PN} != "libiconv" && -n $(ls "${D}"/usr/lib*/charset.alias 2>/dev/null) ]]; then
                einfo "automatically removing charset.alias"
                rm -f "${D}"/usr/lib*/charset.alias
        fi
}

# These are because of
# http://archives.gentoo.org/gentoo-dev/msg_529a0806ed2cf841a467940a57e2d588.xml
# The profile-* ones are meant to be used in etc/portage/profile.bashrc by user
# until there is the registration mechanism.
profile-post_src_install() { bsd-post_src_install ; }
        post_src_install() { bsd-post_src_install ; }

# Another hack to fix old versions of install-sh (automake) where a non-gnu
# mkdir is not considered thread-safe (make install errors with -j > 1)

bsd-post_src_unpack() {
# Do nothing if we don't have patch installed:
if [[ -z $(type -P gpatch) ]]; then
return 0;
fi
local EPDIR="${ECLASSDIR}/ELT-patches/install-sh"
local EPATCHES="${EPDIR}/1.5.6 ${EPDIR}/1.5.4 ${EPDIR}/1.5"
local ret=0
for file in $(find . -name "install-sh" -print); do
if [[ -n $(egrep "scriptversion=2005|scriptversion=2004" ${file}) ]]; then
einfo "Automatically patching parallel-make unfriendly install-sh."
# Stolen from libtool.eclass
for mypatch in ${EPATCHES}; do
if gpatch -p0 --dry-run "${file}" "${mypatch}" &> "${T}/patch_install-sh.log"; then
gpatch -p0 -g0 --no-backup-if-mismatch "${file}" "${mypatch}" \
&> "${T}/patch_install-sh.log"
ret=$?
break
else
ret=1
fi
done
if [[ ret -eq 0 ]]; then
einfo "Patch applied successfully on \"${file}\"."
else
ewarn "Unable to apply install-sh patch. "
ewarn "If you experience errors during install phase, try with MAKEOPTS=\"-j1\""
fi
fi
done
}

profile-post_src_unpack() { bsd-post_src_unpack ; }
post_src_unpack() { bsd-post_src_unpack ; }

Re: [SoC 2009] Gentoo/NetBSD important stuff (!)

Camille Huot — 2009-10-09T15:29:13

Hey Patrice, well done!

If Davide agrees, you may want to update and improve the current
Gentoo/Netbsd Install guide
(http://www.gentoo.org/proj/en/gentoo-alt/bsd/nbsd/doc/gentoo-netbsd.xml)
using your own stuff.

Best regards,

[SoC 2009] Gentoo/NetBSD important stuff (!)

Patrice Clement — 2009-10-09T12:51:42

Hello everybody,

I've been quiet since 1 month, sorry for the delay but it's maily because of
my studies.

Thanks to Alexis Ballier, I've been able to upload my files somewhere on the
web; actually it's hosted on a Gentoo webserver, waiting for somewhere else..

Here is the URL [1]. You'll find:
- A stage3 Gentoo/NetBSD tarball. To install it, you'll have to download NetBSD
  [2] and perform a minimal installation. If you have troubles during
installation, I advise you to read Chapter 3 of NetBSD handbook, "Example
installation", which describes how to do, step by step [3].

- A QEMU image, archived using lzma. This is the quickest way to test my work,
  without performing an installation. To proceed, QEMU has to be installed on
your machine. I let you read documentation on how to proceed to let your VM
surf the web (hint: you must use tun/tap(4) driver), in order to download
ebuilds from it.

Please, feel free to ask questions or to make suggestions, I'll be glad to
answer. I'm currently applying to become a Gentoo developer, to continue
working on that project, I don't want it to die !

Cheers,

Patrice

[1] http://dev.gentoo.org/~aballier/gnbsd/
[2] ftp://ftp4.fr.netbsd.org/pub/NetBSD/NetBSD-5.0/i386/installation/cdrom/
[3] http://www.netbsd.org/docs/guide/en/chap-exinst.html

[SoC 2009] End of SoC 2009: Gentoo/NetBSD project final report

Patrice Clement — 2009-08-24T12:27:56

Hello everyone,

Today is the end of Summer of Code. As soon as I can (= when it'll be possible
to upload this stuff somewhere), I'll release 3 important things to the
Gentoo/BSD community:

- Gentoo/NetBSD ebuilds: as I said, these ebuilds need to be polished, reworked and
  tested. I didn't get a lot of feedbacks on them so please, be soft with me. :P
- gentoo-netbsd-5.0-stage3.tar.bz2: the famous tarball in question. I'll write
  a little wiki to explain installation steps but it should be very easy: perform
a minimal NetBSD install, download the tarball, extract, and enjoy !
- QEMU VM image: for people who don't want to mess with the install, an image
  will be available. I really need to wait for Google to give us a place to
upload our stuff because of the size of the VM (3 GB). Please, be patient. I'm
sorry but currently, I don't have enough place on my VPS. (perhaps Gentoo guys
can provide me some .. ? :D)

I've made some screenshots of this VM during boot process via OpenRC, login,
emerging of an ebuild (nano), .. Take a look at this url to see them:
http://omani.ac/4ap

This project won't die after GSoC; I'd like to continue working on this in
order to make something as mature as Gentoo/FreeBSD project. Yeah, there is a
big amount of work. :)

As usual, don't hesitate to comment my work, I'll be glad to hear you. :)

I'd like to thank Luca Barbato for his support during the half my GSoC; I
probably wouldn't have finished this project without him. I also like to thank
gentoo-bsd mailing list guys for their precious answers. Even if their has been
hard times, it has been a pleasure to work for the Gentoo Foundation. Thank you
all of you !

Cheers, 

Patrice

[SoC 2009] Gentoo/NetBSD ? Of course it runs !

Patrice Clement — 2009-08-19T19:04:45

Heya people,

This is the final GSoC week and we are near from the end: I'll soon release my
work. Yesterday, I've tried to boot a NetBSD machine under QEMU, where my stage
has been extracted and guess what? It boots, I can login into, and even emerge ebuilds!

I've planned to release 3 things to the community:
- my ebuilds: they need to be reworked and to be heavily tested. The only
  feedbacks I get has been from Davide and Luca. Don't hesitate to make
suggestions and comments on my work.
- a stage: currently, the stage contains everything you'll need to run the
  system. Fetch NetBSD 5.0, perform a minimal install and untar the stage on /,
as a normal Gentoo/Linux install.
- a QEMU image: to quickly test my work, without losing time of an
  installation, you can try this image. Simply boot the image and enjoy.

I'm 5% from the end. What are the 5% left ?
A very weird behavior has appeared: on my NetBSD box, when I emerge an ebuild
in the ROOT, as example: nano, ebuild is emerged, chrooting inside the ROOT and
running the binary works fine. No problem with that. So where is the error ?
When I emerge the same ebuild inside the ROOT, it works but each time you
execute this binary, it produces a gmon.out file. Even without using emerge
(downloading sources, ./configure & make), it also produces a gmon.out after
beeing executed. I don't think the problem comes from emerge. But from where ?
GCC comes from NetBSD sources, as same as binutils.
I can read in GCC man page that gmon.out is produced when -pg flag is used.
Here are my CFLAGS & CXXFLAGS variables from /etc/make.conf, in my ROOT directory:
  CFLAGS="-march=i386 -pipe -O2"
  CXXFLAGS=${CFLAGS}
I use the same outside the ROOT (on my NetBSD devel box). I've never used this
-pg flag. Ever !

Here's an example of this weird behavior:

(I emerge nano as an example, but you can emerge whatever you want, it's the
same)

# emerge -v nano
Calculating dependencies... done!

[ebuild   R   ] app-editors/nano-2.1.9 to /gentoo-nbsd/ USE="-debug -justify-minimal -ncurses -nls -slang -spell -unicode" 0 kB
Total: 1 package (1 upgrade), Size of downloads: 0 kB


[...]
 * More helpful info about nano, visit the GDP page:
 * http://www.gentoo.org/doc/en/nano-basics-guide.xml

 * GNU info directory index is up-to-date.

(ok, we're good)
# chroot /gentoo-nbsd /bin/bash
(we're now inside the ROOT)
soc-netbsd ~ # mkdir output
soc-netbsd ~ # cd output/
soc-netbsd output # ls .
soc-netbsd output # nano
(nano opens itself, I close it)
soc-netbsd output # ls .
soc-netbsd output # emerge -v nano
Calculating dependencies... done!

[ebuild   R   ] app-editors/nano-2.1.9 USE="-debug -justify -minimal -ncurses -nls -slang -spell -unicode" 0 kB
Total: 1 package (1 upgrade), Size of downloads: 0 kB


[...]
 * More helpful info about nano, visit the GDP page:
 * http://www.gentoo.org/doc/en/nano-basics-guide.xml

 * GNU info directory index is up-to-date.

soc-netbsd output # ls .
soc-netbsd output # nano
(open/close)
soc-netbsd output # ls .
gmon.out

I've certainly missed something but I still don't get it. If someone have any
clues about this and could give me a hand, that could be nice.

Thanks !

Cheers,
Patrice

Re: [SoC 2009] Ping ? Pong ! News and progress..

Patrice Clement — 2009-08-16T11:14:28

Hi Javier, and thank you for your reply.

Saturday 15 Aug 2009 10:44:27 (-0300), Javier Villavicencio wrote :

That seems to be a huge amount of work. As said before, I'll keep things
"as-is" for now because it just works. This week, I'll focus on problems
solving with PAM and stage building. I'm sorry but I'm a bit in a hurry due to
lack of time. I want to bring something functionnal for everyone. I really want
to continue my work after GSoC, don't doubt on my motivation. You, guys and
devs of this mailing list, were always here to bring me some piece of advices
when I was in trouble.. I want to give a back to you. :)

Cheers!

Patrice

Re: [SoC 2009] Ping ? Pong ! News and progress..

Javier Villavicencio — 2009-08-15T13:44:27

Basically, keep NetBSD versions of those utilities to get NetBSD itself
built successfully, then alias the GNU versions for portage to compile
ebuild successfully.


Ahh good old times. FreeBSD devs did also patch GCC sources to work on
FreeBSD. See http://bugs.gentoo.org/192403 for what it takes to get "our
toolchain" to behave properly on FreeBSD. And that's why I asked some
time ago if you managed to compile the kernel with a gentoo toolchain,
it's a big pain in the back :D.

There might be many things patched, however those in the gcc spec are
key to get a working gcc, and lucky us, the specs don't vary too much,
so you can try a diff between NetBSD gcc specs versus GNU gcc specs to
spot what's missing/required.

From there you can start a "copy" of your sys-devel/gcc ebuild of
choice, and add patches as required, once you get a somewhat working
version the patches can be tested/reviewed by the GNU devs to iron out
what's not required, what's correct, and what's wrong.

And here is where the legal BSD vs GPL battle starts, cause you can't
add BSD copyrighted code into GPL code :D, fun fun.

As a last note, beware that some of the ebuilds in portage may not
compile correctly or at all if they don't use a GNU (gentooized, but gnu
at it's heart) toolchain, which includes the versions of the utils
stated above (m4, sed, awk, etc), binutils, and gcc.

Salu2!,
Javier.

Re: [SoC 2009] Ping ? Pong ! News and progress..

Johan Hattne — 2009-08-14T16:24:12

Dear all;

I did attempt to get NetBSD running with Gentoo on a spare Alpha I had
lying around a couple of years ago.  I have since then moved to where
there's 120V in the wall sockets, and the cheap PSU I used to house the
board cannot deal with it.  I still haven't found the time to buy a new
power supply, and hence the alpha is currently serving as a door stop.


On 08/14/09 06:58, Alexis Ballier wrote:


I seem to recall the NetBSD people introduced new functionality into
OpenPAM without which NetBSD won't work.  I've haven't been following
NetBSD or OpenPAM very closely since I gave up on my alpha, so I'm not
sure how much help my old patch would be.  But you can make your own!
Just make a diff between the sources of sys-auth/openpam and
corresponding sources in the NetBSD tree.


On 08/14/09 06:58, Alexis Ballier wrote:


I vaguely recall having problems here as well.  As Alexis points out, I
think it turned out to be related to OpenSSL.  It may be worth checking
_very_ carefully that your openssl build went fine.  As I remember it,
it all seemed to compile OK, but due to something being misdetected or
something in one of those early perl scripts a lot of it was actually
garbage.

If you really want, I'll try to get the alpha running again.  Maybe you
could help me get it to boot to multiuser?  I never did figure out how
to get the rc system and baselayout and whatnot to agree with NetBSD (I
particularly remember that there was no fuser on NetBSD and porting it
from FreeBSD was not trivial).

// Cheers; Johan

Re: [SoC 2009] Ping ? Pong ! News and progress..

Patrice Clement — 2009-08-14T14:50:22

Friday 14 Aug 2009 13:15:07 (+0200), Alexis Ballier wrote :

Shit happens. But thanks for your help. :)
I'm gonna ping you on IRC.

Ok, I'll correct that.

Using OpenRC makes rcorder useless: it's used in /etc/rc script.

I didn't know that, and it's great if "g" suffix is added.
But Portage needs some GNU tools like GNU sed: some scripts/patches
don't work if you don't use it. Same thing about awk:
I read ebuild.sh source code and "gawk" command is used. I'll correct this but
I'll still install them in my stage.

On my "correct" list.

Yes, NetBSD has also his own version of uuidgen and libuuid. On my list.

*,
which means: we should talk about this on IRC. I'll give you errors logs. It's
mainly compilation errors.

I'm gonna rename it.

NetBSD devs seem to have patched GCC sources to suit to NetBSD. I maybe writing
a big mistake. But I have tried every GCC's ebuilds and none of them have
worked. Even by downloading GCC sources on gcc.gnu.org website, ./configure and
make, it simply doesn't work. That's why I've created an ebuild..

Yes, it should work. In fact, it's because I haven't yes packaged my ebuilds
that I've used patches. To compile and install one of my ebuilds, you'll need
to download sources via CVS. My ebuilds are just a frontend to /usr/src. I know
it's not the right way to do: Davide seemed to be fine with that, but Luca
explained me how to proceed. For the moment, I'll stay with this configuration
because it "works". But, of course, I'll correct this after GSoC.

~ grep "EVP_PKEY_decrypt" /usr/include/openssl/evp.h
int EVP_PKEY_decrypt(EVP_PKEY_CTX *ctx, unsigned char *out, size_t *outlen, const unsigned char *in, size_t inlen);

I've tried to compile Kerberos 5 1.7 this morning and it works without changing
any ./configure flags. The only error comes from PKINIT: it is an extension
to Kerberos that adds public key cryptography. Do we really need it ?
Thus, if I remove it from Makefile, compilation works fine.

Yes. The problem comes from "passwd" program which segfaults. I need to dig on
this more. I can provide a .core if you want.

Anyway, thanks again for your help. :)

Cheers!

Patrice