[gentoo-user] my xorg-server 1.6 seems a bit unstable - wh..

I've been having some issues myself (mythtv manages to mess up my screen
royally, I need to switch to a text virtual console and back to X to
make the screen legible after viewing a TV-recording in myth). I just
noticed that the version of the headers for xcb do not match the version
on libxcb itself. Could this be the cause of my (and Daves) problems? I
have libxcb 1.4, but headers version 1.5, no 1.4 available for the
headers. (running a new sync of my portage tree now, just in case I've
missed something)
Viz:

1:medisin ~ # eix xcb
[I] x11-libs/libxcb
Available versions: 1.0 1.1 ~1.1.90.1 1.4-r1 {debug doc selinux}
Installed versions: 1.4-r1(22:28:37 10/11/09)(-debug -doc -selinux)
Homepage: http://xcb.freedesktop.org/
Description: X C-language Bindings library

[I] x11-libs/xcb-util
Available versions: 0.3.3 ~0.3.4 ~0.3.5 0.3.6 {debug test}
Installed versions: 0.3.6(18:11:17 10/08/09)(-debug -test)
Homepage: http://xcb.freedesktop.org/
Description: X C-language Bindings sample implementations

* x11-misc/xcb
Available versions: 2.4 {motif}
Homepage: http://www.goof.com/pcg/marc/xcb.html
Description: Marc Lehmann's improved X Cut Buffers

[I] x11-proto/xcb-proto
Available versions: 1.0 1.1 ~1.2 1.5
Installed versions: 1.5(05:16:09 10/06/09)
Homepage: http://xcb.freedesktop.org/
Description: X C-language Bindings protocol headers

Found 4 matches.

Alan McKinnon wrote:
> On Tuesday 13 October 2009 21:04:54 Denis wrote:
>
>> Perhaps I should try this procedure on my home gentoo box first.
>>
>> Here are my USE flags:
>>
>> X aac aim alsa ao bash-completion bdf branding bzip2 cairo cdda cddb
>> cdparanoia cdr cpudetection djvu emacs encode exif firefox flac
>> foomaticdb gif glut graphics gs gtk ieee1394 imagemagick imlib java
>> javascript jbig jpeg jpeg2k kpathsea lame latex ldap libnotify libwww
>> lm_sensors mad mmx mmxext mng motif mp3 mpeg mplayer msn openal opengl
>> oss pdf plotutils png ppds preview-latex pstricks qt3support qt4
>> quicktime rtc samba science sdl sensord smp sndfile spell sse sse2
>> svga swat tcl tetex tiff tk truetype tta usb wavpack winbind wmp xcb xine
>> xml yahoo -dri -isdnlog -pppd
>>
>
> At a minum you'll need to remove
>
> X
> cairo
> firefox
> gtk
> motif
> opengl
> pdf
> qt3support
> qt4
> truetype
> scb
>
> There will be others, that's just obvious ones from a quick scan through. Also
> remove similarly obvious stuff from package.use
>
> The as Dale said, manually edit your world file to remove everything that is
> obviously X related. But first, verify that "emerge -avuND world" want to
> merge nothing, otherwise --depclean will complain.
>
> Then the fun starts:
>
> emerge -a --depclean and let it do whatever it wants.
> emerge -pvuNDt world and look closely at what X stuff it wants to pull in.
> See if it's a dependency thing or a USE flag thing, and deal with it.
> Eventually, emerge -pvuND world will not want to remerge anything, only
> rebuild stuff due to changed USE flags.
> --depclean again, rinse and repeat as many times as you have to (probably
> lots).
>
> When you have done all this, and it will take a day or three, you will no
> doubt realise that it was not worth the effort and you should have just
> reinstalled without X, or fixed the underlying problem.
>
> Excising X is like downgrading to stable from unstable - not worth the effort.
>
> Sorry, but them's the breaks
>
>
>> When "world" is edited, should I leave make.conf be or do I also need
>> to remove all X-related USE flags from there?
>>
>
> Remove the flags from make.conf and also from package.use
>
>
>
>

And if you run into a snag, try this option with emerge: --with-bdeps y
I actually have that in make.conf for mine.

As mentioned by Alan, I really doubt this is going to do much if any
good. After you did a emerge -e world, that is pretty much it.

Let us know what blows up.

Dale

On Tuesday 13 October 2009 21:04:54 Denis wrote:
> Perhaps I should try this procedure on my home gentoo box first.
>
> Here are my USE flags:
>
> X aac aim alsa ao bash-completion bdf branding bzip2 cairo cdda cddb
> cdparanoia cdr cpudetection djvu emacs encode exif firefox flac
> foomaticdb gif glut graphics gs gtk ieee1394 imagemagick imlib java
> javascript jbig jpeg jpeg2k kpathsea lame latex ldap libnotify libwww
> lm_sensors mad mmx mmxext mng motif mp3 mpeg mplayer msn openal opengl
> oss pdf plotutils png ppds preview-latex pstricks qt3support qt4
> quicktime rtc samba science sdl sensord smp sndfile spell sse sse2
> svga swat tcl tetex tiff tk truetype tta usb wavpack winbind wmp xcb xine
> xml yahoo -dri -isdnlog -pppd

At a minum you'll need to remove

X
cairo
firefox
gtk
motif
opengl
pdf
qt3support
qt4
truetype
scb

There will be others, that's just obvious ones from a quick scan through. Also
remove similarly obvious stuff from package.use

The as Dale said, manually edit your world file to remove everything that is
obviously X related. But first, verify that "emerge -avuND world" want to
merge nothing, otherwise --depclean will complain.

Then the fun starts:

emerge -a --depclean and let it do whatever it wants.
emerge -pvuNDt world and look closely at what X stuff it wants to pull in.
See if it's a dependency thing or a USE flag thing, and deal with it.
Eventually, emerge -pvuND world will not want to remerge anything, only
rebuild stuff due to changed USE flags.
--depclean again, rinse and repeat as many times as you have to (probably
lots).

When you have done all this, and it will take a day or three, you will no
doubt realise that it was not worth the effort and you should have just
reinstalled without X, or fixed the underlying problem.

Excising X is like downgrading to stable from unstable - not worth the effort.

Sorry, but them's the breaks

> When "world" is edited, should I leave make.conf be or do I also need
> to remove all X-related USE flags from there?

Remove the flags from make.conf and also from package.use



--
alan dot mckinnon at gmail dot com

Ordinarily, I would also point to I/O errors, but not this time. I
have had this same version of Mathematica, 5.2, for *years*, on this
same machine, over several kernels and several Xorg upgrades, and it
was rock-solid. Not one X hang or crash. Ever. No error messages or
I/O warnings ever. The only thing I'd ever see in the console would
be GTK+ errors from other programs. Something about this latest X
upgrade is making Mathematica unhappy, either because some previously
supported format that worked fine with it has been altered or
something else in that general regard. This version being so outdated
- that isn't a surprise. It may not be a bug, so much as an
incompatibility, perhaps

I think the only thing I can do at this point that probably will fix
this is to go to stable arch, but I am not willing to do that because
I like my box the way it is now It's great and efficient for all
other things. Before I can get a local Mathematica upgrade, I will
have to resort to using it via X forwarding off the central server.

Denis wrote:
> Ordinarily, I would also point to I/O errors, but not this time. I
> have had this same version of Mathematica, 5.2, for *years*, on this
> same machine, over several kernels and several Xorg upgrades, and it
> was rock-solid. Not one X hang or crash. Ever. No error messages or
> I/O warnings ever. The only thing I'd ever see in the console would
> be GTK+ errors from other programs. Something about this latest X
> upgrade is making Mathematica unhappy, either because some previously
> supported format that worked fine with it has been altered or
> something else in that general regard. This version being so outdated
> - that isn't a surprise. It may not be a bug, so much as an
> incompatibility, perhaps
>
> I think the only thing I can do at this point that probably will fix
> this is to go to stable arch, but I am not willing to do that because
> I like my box the way it is now It's great and efficient for all
> other things. Before I can get a local Mathematica upgrade, I will
> have to resort to using it via X forwarding off the central server.
>
>
>

I forgot to mention this earlier. The way I found out it was a I/O
error for me was this command. hdparm -Tt /dev/hda && hdparm -Tt
/dev/hda . I kept repeating that until it starting giving me errors.
For mine it was after the third pass. Something to keep in mind, it
could be something wrong with the driver in the kernel you are using.
It could be any number of things like this. It could be that it is a
mismatch as you mentioned between Mathematica and some other support
program such as gtk, python or any other number of things.

Me, I would hop on a mailing list for that software and ask them what
strace command you can run and then send them the results. It may be
that they can make some sense of the error and recommend a fix or even a
downgrade of some other package.

You may also want to use the q command to see what was upgraded since it
last worked. That might point to something too. I think it is the q
command that does that.

Dale

> Have you tried "strace'ing" mathematica to see why it dies?
>
> BillK

I ran an strace, but I can't make sense out of it. There's a bunch of
polling and trying to access unavailable resources, from what I can
see... The file is huge, but here is the tail end of it:

=====================================
poll([{fd=11, events=POLLIN|POLLPRI}, {fd=21, events=POLLIN|POLLPRI}],
2, 0) = 0 (Timeout)
gettimeofday({1255560657, 709869}, NULL) = 0
gettimeofday({1255560657, 709891}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
gettimeofday({1255560657, 709945}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
poll([{fd=11, events=POLLIN|POLLPRI}, {fd=21, events=POLLIN|POLLPRI}],
2, 0) = 0 (Timeout)
gettimeofday({1255560657, 710025}, NULL) = 0
gettimeofday({1255560657, 710046}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{";\0\5\0o\5 \1\0\0\0\0\0\0\0\0\202\4$\0;\0\5\0!\0
\1\0\0\0\0\0"..., 2204}, {NULL, 0}, {""..., 0}], 3) = 2204
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{"=\0\4\0|\5 \1\2\0\16\0\v\0)\4>\4\7\0w\5 \1|\5 \1.\3
\1\0"..., 100}, {NULL, 0}, {""..., 0}], 3) = 100
poll([{fd=10, events=POLLIN}], 1, -1) = 1 ([{fd=10, revents=POLLIN}])
read(10, "\6\0pa\23\4\t\1I\2\0\0|\5
\1\0\0\0\0\6\5\315\0\6\0`\0\0\1\1\0\16"..., 4096) = 96
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
gettimeofday({1255560657, 718652}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{";\0\5\0o\5 \1\0\0\0\0\0\0\0\0\202\4$\0;\0\5\0!\0
\1\0\0\0\0\0"..., 640}, {NULL, 0}, {""..., 0}], 3) = 640
poll([{fd=10, events=POLLIN}], 1, -1) = 1 ([{fd=10, revents=POLLIN}])
read(10, "\6\0\335a\34\4\t\1I\2\0\0|\5
\1\0\0\0\0\6\5\314\0\6\0_\0\0\1\1\0\1"..., 4096) = 64
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{"8\0\4\0o\5 \1\4\0\0\0\0\0\0\0>\4\7\0\3\6 \1\202\5 \1o\5
\1\0"..., 48}, {NULL, 0}, {""..., 0}], 3) = 48
poll([{fd=10, events=POLLIN}], 1, -1) = 1 ([{fd=10, revents=POLLIN}])
read(10, "\16$\364a\202\5
\1\0\0>\n\0\371\34\10\361\207\t\10\320n1\n\332\0o\0}\3~\0\1"..., 4096)
= 64
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
gettimeofday({1255560657, 723910}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
poll([{fd=11, events=POLLIN|POLLPRI}, {fd=21, events=POLLIN|POLLPRI}],
2, 0) = 0 (Timeout)
gettimeofday({1255560657, 723992}, NULL) = 0
gettimeofday({1255560657, 724014}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
gettimeofday({1255560657, 724068}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
poll([{fd=11, events=POLLIN|POLLPRI}, {fd=21, events=POLLIN|POLLPRI}],
2, 0) = 0 (Timeout)
gettimeofday({1255560657, 724148}, NULL) = 0
gettimeofday({1255560657, 724181}, NULL) = 0
poll([{fd=10, events=POLLIN}, {fd=11, events=POLLIN|POLLPRI}, {fd=21,
events=POLLIN|POLLPRI}], 3, 0) = 0 (Timeout)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{";\0\5\0o\5 \1\0\0\0\0\0\0\0\0\202\4$\0;\0\5\0!\0
\1\0\0\0\0\0"..., 1536}, {NULL, 0}, {""..., 0}], 3) = 1536
read(10, 0x9d67130, 4096) = -1 EAGAIN (Resource
temporarily unavailable)
poll([{fd=10, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=10, revents=POLLOUT}])
writev(10, [{"=\0\4\0|\5 \1\2\0\16\0\v\0)\4>\4\7\0w\5 \1|\5 \1.\3
\1\0"..., 100}, {NULL, 0}, {""..., 0}], 3) = 100
poll([{fd=10, events=POLLIN}], 1, -1) = 1 ([{fd=10, revents=POLLIN|POLLHUP}])
read(10, ""..., 4096) = 0
close(2) = 0
dup2(20, 2) = 2
write(2, "XIO: fatal IO error 11 (Resource"..., 79) = 79
write(2, " after 90691 requests (90613"..., 77) = 77
exit_group(1) = ?
=====================================

Does this make any sense to anyone?

On 14 Oct 2009, at 17:50, Denis wrote:
> ... While some folks
> here probably think this attempt foolish, I still think it was worth a
> try, and I learned some new things, which is always good.

I don't think this attempt foolish - but I guess you're getting
desperate.

Having performed `emerge -e world` I would come to the conclusion that
the problem is simply Mathematica, with some current libraries. I
would probably want to try a fresh install of Gnetoo - on a different
box or hard-drive - to prove the point. But I think I would be
concluding shortly that it's an upstream bug.

Stroller.