gmane.linux.lfs.beyond.support

Re: MesaLib take about 18 SBU on my box.

John Burrell — 2012-05-24T23:53:22


Yes, I think it is some env variable that is not set on my machine, but is on yours.

When I make mesalib the first time, it sets up the environment, so when I ctrl^Z 
and make again the variable(s) are now correctly set and it builds. It must be
something like this anyway but I still don't know what to set before I try and build it.

I checked if it was a chroot problem, but it isn't. It behaves exactly the same way whether
chrooted, or logged in to LFS.

jb.

Re: MesaLib take about 18 SBU on my box.

Bruce Dubbs — 2012-05-24T19:14:02


This is not needed for me.  I think it is some sort of system 
configuration problem.

I have makedepend in /opt/xorg/bin.  What I did was:

mv makedepend makedepend.bin

I then created a new makedepend:

#!/bin/bash
echo "`date +%X.%N` Start"  >> depend.dates
makedepend.bin $< at >
echo "`date +%X.%N` End"  >> depend.dates

---------

chmod 755 makedepend


And reran the make portion of mesalib.   The mesa scripts redirect 
stdout and stderr to /dev/null.

I then ran:

find -name depend.dates  -exec cat {}  \; |sort

You can see from the results below that makedepend only should take a 
fraction of a second.

   -- Bruce

14:06:23.029787971 Begin
14:06:23.048103509 End
14:06:25.662653374 Begin
14:06:25.691832992 End
14:06:25.695225702 Begin
14:06:25.725778424 End
14:07:15.530298089 Begin
14:07:15.689573695 End
14:07:15.693863551 Begin
14:07:15.859673535 End
14:09:52.234531014 Begin
14:09:52.280427304 End
14:10:05.461738450 Begin
14:10:05.524091781 End
14:10:39.438888202 Begin
14:10:39.478076028 End
14:10:47.19

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-24T17:44:29

2012/5/24 John Burrell <john_burrell< at >hotmail.com>:
This way work, but will this break somethings ?
Is this a bug of MesaLib or makedepend ? We'd better report it to upstream.

Re: MesaLib take about 18 SBU on my box.

John Burrell — 2012-05-24T15:12:36



----------------------------------------

I have this problem too. I haven't found a solution to it yet.

As soon as you see it go to:

running  /usr/bin/makedepend

you should ctrl^Z and and then type make again.

It will complete in the normal time.

HTH

jb.

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-24T14:28:16

2012/5/24 Ken Moffat <zarniwhoop< at >ntlworld.com>:

I have not use llvm

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-24T14:27:35

I've tried many way, it still need very long time.
It spend most time on doing makedepend.

Re: MesaLib take about 18 SBU on my box.

Ken Moffat — 2012-05-24T14:22:45

 It would have been better if I checked what I actually do, instead
of relying on my memory :)  The only place in my scripts to build
Xorg where I run ldconfig is for llvm.

ĸen

wpa_supplicant-1.0 fails to compile

John Burrell — 2012-05-24T14:21:53

wpa_supplicant-1.0 fails with:

../src/drivers/driver_nl80211.c:25:31: fatal error: netlink/genl/genl.h: No such file or directory
compilation terminated.

I had to include the libnl3 headers location:

CFLAGS="-I/usr/include/libnl3" make BINDIR=/sbin LIBDIR=/lib

for it to work.

jb.

Re: MesaLib take about 18 SBU on my box.

Bruce Dubbs — 2012-05-24T02:49:09


Hmm.  1G is not much any more. You are probably getting into swap.  My 
SBU time on my i686 pentium4 is 132.5 seconds, but even though it's 
seven years old, it is 3.2GHz.

   -- Bruce

Re: MesaLib take about 18 SBU on my box.

Armin K — 2012-05-24T01:32:28

Pentium 4 HT with 1GB of DDR RAM. I can't remember, it is nearly a year 
since I measured SBU, and I upgraded my memory not so long ago.

Re: MesaLib take about 18 SBU on my box.

Ken Moffat — 2012-05-23T21:39:18

 I think you mean that you did not run ldconfig after each library
was installed ?  If so, any error would happen when ld failed to
find the library, and you would get a build failure.  So, I doubt
this is the problem (although it is, theoretically, possible that
xorg-server could fail to link).

 Either way, running ldconfig when you encounter a failure to link
will pick up all the now-existing libraries it did not know about.

 Certainly, we used to have to run ldconfig for many (or all) of the
libraries when monolithic xorg (6.8 ?) was first broken into the
modular version, and I still do that.

 So, I need to remove the 'ldconfig' steps on my next build (I'm now
on 7.7-rc, so *if* they are no longer necessary for me, they might
still be needed in 7.6) and test this.  Thanks for rousing my
interest, I'm *so* glad I said I had no plans to upgrade anything
else on BLFS at the moment ;-)

 Whatever, I hope your next build of Mesa is quicker!

ĸen

Re: [Fwd: Error in BLS Python-2.7.3]

Ken Moffat — 2012-05-23T21:29:16

 A quick look at my local copy suggests it is in option tags, not
parameter tags, and the explanation says 'if you have installed...'

 If I'm correct, Marcos might wish to note that there are a lot of
optional explanations for switches we don't include in the book's
configure instructions.  On firefox, the parameter explanations
(things we mentioned above) are rendered differently from the option
explanations (other things that someone decided to explain).  ISTR
one is regular text, the other italic, but I'm sitting in front of a
tty at the moment).  For lynx, obviously they will look alike.

ĸen

[Fwd: Error in BLS Python-2.7.3]

Bruce Dubbs — 2012-05-23T20:49:41

-------- Original Message --------
Subject: Error in BLS Python-2.7.3
Date: Wed, 23 May 2012 17:00:53 -0300
From: Marcos Yoshito Kasahara <kasahara< at >ig.com.br>
To: bdubbs< at >linuxfromscratch.org

Dear Bruce

I think I have found a mistake in Python-2.7.3 (
http://www.linuxfromscratch.org/blfs/view/svn/general/python2.html),
in the line "Command Explanations" you mention "--with-system-ffi", but in
the actual command line there is no
"--with-system-ffi".

sed -i "s/ndbm_libs = \[\]/ndbm_libs = ['gdbm', 'gdbm_compat']/" setup.py &&
./configure --prefix=/usr --enable-shared &&
make

Thanks

Best Regards

Marcos Kasahara

Re: MesaLib take about 18 SBU on my box.

Bruce Dubbs — 2012-05-23T20:47:45


Ouch.  Are you building on a 386?  :)

   -- Bruce

Re: MesaLib take about 18 SBU on my box.

Ken Moffat — 2012-05-23T20:27:33

 Realised I'd overlooked the 18 SBU in your title.

 On my current LFS-svn system (end of April), a true SBU is 209
seconds (i.e. remeasured after booting it, using make -j1).

 The book's llvm was recently extended to optionally include
clang-3.0, I still build llvm without that and the way we used to,
so I suppose your build might have decided to use clang instead of
gcc for the C compiler ?

 Anyway, on my current system (Phonon x86_64, cpufreq ondemand
controller running variously at 800,2200,2700,3400 MHz, recent
7200rpm SATA drive) using make -j1 and with all the gallium drivers,
configure takes 5 seconds, make takes 7 minutes 16, DESTDIR install
takes 8 seconds (all rounded to the nearest second).  So for me,
this time, that is 2.1 SBU (and 183MB).

 In passing, I notice that my current test build on my i3 (host
LFS-7.1, make -j3 with the original SBU also using -j3 and not
recalculated, with a multiplicity of cpu frequencies up to 3300)
the recorded SBU was 3.9.  Strange, but not worth me worrying.

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-23T19:36:38

May be this reason, I lost "ldconfig" in Xorg Libraries.
I'll try it tomorrow, and feed back

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-23T19:28:01

2012/5/24 Bruce Dubbs <bruce.dubbs< at >gmail.com>:
nothings else is running while compiling Mesalib

Re: MesaLib take about 18 SBU on my box.

xinglp — 2012-05-23T19:26:01

2012/5/24 Armin K <krejzi< at >email.com>:

About 90 seconds 1 SBU on my machine.
The Mesalib-8.0.2 take about 27 minutes on my machine.
My machine:
CPU: SL9400 2 cores
MEM: 2G
DISK: 5400 2.5" ST9250827AS

Re: MesaLib take about 18 SBU on my box.

Armin K — 2012-05-23T18:55:03

How much is that in minutes and how much is 1 SBU on your machine?

It takes arround 30 minutes here, where 1 SBU is arround 8 minutes (I 
think).

Re: MesaLib take about 18 SBU on my box.

Ken Moffat — 2012-05-23T18:38:54

 SBUs are only a rough guide.  They get less accurate when older
toolchains are involved (e.g. if your host had an older gcc minor
version than what was in the LFS version you built, your original
SBU will be small because each new version of gcc takes longer to
compile programs, so all your later LFS and BLFS SBUs will be
bigger.  Similarly, if you are building on an older version of LFS
your BLFS SBUs will ofteny be quicker.

 In this particular case, I suspect Mesa might have been measured
using LFS-7.0, not 7.1 (we don't normally update measurements when we
update a BLFS checked/built tag, unless things are clearly wrong).

 Also, that measurement is presumably for the limited set of gallium
drivers shown in the book's configure command.  Those of us not using
nvidia hardware will probably wish to enable other, or all, the
gallium drivers (or omit nouveau).

 In my case I build all the drivers on my current hardware (I've got
big disks now, I can afford the space and I'd like to know if any
of them fail

Re: MesaLib take about 18 SBU on my box.

Bruce Dubbs — 2012-05-23T18:31:20

That's a little quick compared to my system.  For MesaLib-7.11.2 I have 
3.349 SBU and for MesaLib-8.0.2 I have 3.886 SBU.

The SBU values are only meant as an approximate figure.

18 SBU is quite long though.  How much time is 1 SBU on your system? 
Was there anything else going on?

   -- Bruce