LM | 9 May 2013 13:30
Picon

Re: lfs-chat Digest, Vol 1328, Issue 1

Aleksandar Kuktin wrote:
I still think that X without a desktop environment and with one or two
small graphical libraries is better.

I'm beginning to think that way as well.  Our local Linux users group was demoing a Raspberry Pi.  They couldn't hook it directly to a monitor.  Without X, they wouldn't have been able to view it on a monitor via the network.

Would be curious what small graphics or GUI libraries would be recommended for such a system.  At present, I'm trying to concentrate on SDL, ncurses/pdcurses and FLTK.  Am currently investigating if I can find enough replacements for programs I commonly use that only require lighter weight GUI libraries.  For instance, I work with audacity a lot.  Some other possibilities might be apcstudio (FLTK) or rezound (Fox Toolkit), but not sure if they offer all the same editing capabilities. 

By the way, another non-X option besides Wayland is to use the framebuffer.  There are various ports of common GUI libraries that work with DirectX and there are some interesting tools that work directly with framebuffer.

Sincerely,
Laura

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
LM | 5 May 2013 17:17
Picon

lightweight library and utility alternatives

Just wondering if anyone's using any more lightweight alternatives on
their systems rather than the usual standards.  For instance, busybox
or toybox instead of the usual GNU tools, gettext-tiny instead of
gettext, pkg-conf instead of pkg-config.  If you're using any
alternative libraries or utilities as replacements, would be curious
to hear what you're using.  Thanks.

Sincerely,
Laura
http://www.distasis.com/cpp
--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Badoo | 4 May 2013 19:58
Favicon

★ Read your message before it gets deleted!

Read the message that Dominic Ringuet left for you before it gets deleted!



Read your message



This email is part of our delivery procedure for the message sent by Dominic Ringuet. If you have received this email by mistake, please ignore it. The message will be deleted soon.

Have fun!
The Badoo Team

You have received this email from Badoo Trading Limited (postal address below). If you do not wish to receive further email communications from Badoo, please click here to opt out.
Badoo Trading Limited is a limited company registered in England and Wales under CRN 7540255 with its registered office at Media Village, 131 - 151 Great Titchfield Street, London, W1W 5BB.

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
Kenno Han | 30 Apr 2013 08:42
Picon

Live CD

Just wondering, what happen to the LFS LiveCD project?

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
Alice Wonder | 29 Apr 2013 11:15
Favicon

init script management

I'm very tempted to try to write something similar to chkconfig but in
Python, primary motive for me is something to have in RPM scriptlets.

But before I do, is there already something? Well, obviously there's
chkconfig but the two versions I know of are compiled (overkill) and a
perl version that apparently is overly complex (I think part of
SuSE ??).

Also, before I do, does it look like LFS is sticking with SystemV init
(my preference) opposed to systemd? I prefer SystemV init but that's
mostly because lately my short term memory isn't so good, and thus it's
hard to learn a new system ;)

If I do write my own, my goal is to try to make it more or less
universal (no need to make special headers in the init script) but have
it potentially use headers (like the dependency part of LFS init script
headers) if they exist.

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Kenno Han | 26 Apr 2013 20:59
Picon

login manager

It would be interesting to have a login manager in LFS.
I bet many people will need this feature. SLiM would be a nice choice.
It is lightweight and easy to use. The installation is pretty straightforward.
I myself has succeeded building SLiM (simple login manager) on my LFS-7.0 system.

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
Eleanore Boyd | 24 Apr 2013 22:41
Picon
Gravatar

LFS in EPub format

Would anyone be interested in downloading .epub versions of the LFS book? I'm already working on converting the older ones into .epubs out of boredom and lack of anything better to do.


Eleanore Boyd
--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
Alice Wonder | 14 Apr 2013 04:09
Favicon

My experience with rpm and lib64

Round One - I did LFS pretty much by the book and installed the needed
dependencies by the LFS book - lua and elfutils are optional but not in
blfs so I didn't build them.

Python 2 is in the book but make test fails a few tests, probably
because of lack of network - but rpm with python support fails to
compile. I'll investigate later.

Before I started bootstrapping I did something hackish and removed the
{,/usr}/lib64 symlinks, moving the lib to lib64, and then moving none
lib64 stuff (like locale) back. It seemed to work.

The configure system of rpm appears to be broken. It does not detect
that I am linux, but instead uses none. However, config.guess does
correctly identify me as linux. Output of config.guess is same as on any
linux distribution. The compiled binary however does know I'm linux.

So the result is rpmbuild looks for linux macros that aren't there, and
very basic macros like %{_arch} are not defined.

cd /usr/lib/rpm/platforms
cp -ar x86_64-none x86_64-linux
cp -ar noarch-none noarch-linux

fixes that issue.

rpmbuild though then failed to build shared libraries

editing /usr/lib/rpm/macros

and changing

%_host                  x86_64-pc-none
to
%_host                  x86_64-pc-linux

and

%_host_os               none
to
%_host_os               linux

seemed to fix that. Now it would build shared libraries, but it could
not extract the dependency or provides information, it was missing the
tools to do that.

Apparently elfutils is not optional, and I wondered if not having it was
what caused the platform mis-detection during configure.

So I installed elfutils, it's install file indicated it was a
simple ./configure && make && make install.

Installing elfutils however broke the compiler. And that is exactly why
I want a proper package management system. My guess is elfutils install
replaced something in the tool chain.

Round 2 - Fortunately I had a second LFS root and scripts for building
LFS (CH5 and CH6) - so I started over from within that root, but this
time modifying the scripts to use {,/usr}/lib64 from the start. That way
it would be cleaner.

There were a few issues:
glibc - the localedef program wants to use /usr/lib64/locale which is
wrong. I did not see an obvious configure switch to make it do the right
thing, probably need to run a post configure sed on a makefile but for
now I manually moved the directory and made a symlink so localedef used
by other programs can find it.

udev - the custom lfs makefile needed a few edits for lib64, I have a
patch if anyone interested, seems to work just fine.

libkmod - it put the libraries in the right place and the .pc file was
correct but it installed into /usr/lib/pkgconfig instead
of /usr/lib64/pkgconfig.

One package from blfs - either nspr or nss (nss I think) put its
libraries in wrong place but it did not have a configure script, so I
suspect a sed will fix that in the future. I manually moved them.

This time I did not build python, I will build python inside of rpm once
I have the system bootstrapped. That will make it easier to try
different configure switches to see if I can get rpm to build against
it. I also did not build elfutils. First I'm going to bootstrap the
system using rpms that don't have the automated dependency stuff, that
way I can build elfutils inside rpm and it will be easier to see what it
conflicts with that broke the compiler, and I can remove it if need be.

But anyway, now I'm in a clean LFS 7.3 with a separate {,/usr}/lib64 and
at least a primitive but working rpm. Tested by building the Fedora 18
lua rpm (it's not in blfs and Fedora has some nice patches that make it
easier to get as a shared library) - it built and installed perfectly :)

So now I'm busy writing spec files, test building them in Fedora 18
(where I can use rpmlint to find issues - mostly documentation that
isn't utf8, iconv usually fixes them - often it's a changelog or old man
page)

nspr and nss are first since one of them (nss I think) wasn't a clean
install into lib64, then I'll go in the order of the book - hopefully
figuring out how to get localedef to use /usr/lib/locale

So - sorry for the length, but I don't have a blog ;) and I thought
someone might be interested.

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Ken Moffat | 11 Apr 2013 23:40

opaque types in C

 Anyone looking at blfs-dev will have noticed I've been having fun
with xine-ui [ Open -> File, via the right mouse button, not working
in 0.99.7 ].  I'll be talking to upstream, but I'm still at a loss
to understand what has happened.

 There is status attached to xine streams.  This is manipuliated by
the xine engine in xine-lib.  For users such as xine-ui it is an
incomplete data type and therefore I can't print its value in
xine-ui files (although I can test it against the known values).

 So, nothing in xine-ui can change this opaque data, right ?

 What is baffling me is that the first change where the status is
altered only moves logo files to a different source directory.

 But after instrumenting the code to find out where it was failing,
I find that all tested versions from that changeset to the current hg
version only work in 'Open -> File' if I change a test from
'== XINE_STATUS_STOP' to '== XINE_STATUS_PLAY' (and doing this
doesn't seem to affect how they work when a file is passed to xine,
or if the MRL browser is used to select a file).

 But this seems impossible.  Anyone recall anything that ever looked
at all like this problem ?  I think I'm heading for breakfast at
Milliways ¹

 To be clear, xine-lib is the same version for all of this, 0.99.6
works fine, as do all tested versions before that first problematic
changeset.

ĸen

¹ : if you don't understand the phrase, see
http://www.urbandictionary.com/define.php?term=Milliways&defid=820286
-- 
das eine Mal als Tragödie, das andere Mal als Farce
--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
Alice Wonder | 10 Apr 2013 11:15
Favicon

rpm in LFS, /lib64 and /usr/lib64

Hello list,
This is neither LFS nor BLFS so I think this list is best?

Built a LFS 7.3 system and first thing I did was build the necessary
dependencies (according to BLFS instructions for them) to build the RPM
Package Manager.

What I would like to do, and I'll need to do it before I start
bootstrapping the system in RPM, is delete the /lib64 and /usr/lib64
symlinks and create directories in their place so that 64 bit shared
libraries can go into them.

This will likely break building at least temporarily.

Plan to fix building is to initially create symlinks for the libraries
in {,usr}/lib pointing to {,usr}/lib64 and a symlink in /usr/lib for
pkgconfig pointing to ../lib64/pkgconfig

eventually as I rebuild everything in RPM the pkg-config files should
correctly point to 64-bit specific library paths and the pkgconfig
symlink and the library symlinks can be deleted.

libtool .la files - I'll back them up somewhere but I really would like
to get rid of them, Fedora world got rid of them long time ago, they
cause problems (as does rpath)

other than pkg-config files, is there anything else anyone can think of
that might end up biting me?

I have no intent of trying to go multilib but if there ever is a need I
would like it to be as painless as possible and 64-bit specific
libraries and data files (kernel modules excepted) in 64-bit specific
directory makes sense because it is already being done that way on other
multilib RPM based systems.

Thanks for comments / suggestions.

--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Ken Moffat | 31 Mar 2013 06:09

cpufreq on SandyBridge in 3.9

if anybody cares (;-)

 In the 3.9-rc kernel there is a new cpufreq driver for Intel
SandyBridge CPUs, enabled by CONFIG_X86_INTEL_PSTATE.  The help
says:

 This driver provides a P state for Intel core processors.
 The driver implements an internal governor and will become
 the scaling driver and governor for Sandy bridge processors.

 When this driver is enabled it will become the perferred
 scaling driver for Sandy bridge processors.

 Since I have a SandyBridge processor on this machine, I enabled
this setting.  The first thing I noticed was that my own cpufreq
initscript failed - I set 'ondemand' as the .config default, and in
the initscript I set that for each core.  But this driver only
provides powersave or performance.  A workaround to set the old
acpi driver is to add intel_pstate=disable to the grub command line.

 After fairly extensive testing, I can confirm that the new driver
is reasonable - my .config default of 'ondemand' brings up the
'powersave' setting which is not very useful - a setting slightly
slower than with powersave in the acpi driver, probably good for
battery life but painful when compiling :)

 The 'performance' setting, OTOH, is good - with the acpi driver
'performance' stays at full speed (3300 MHz in my case), but with the
new P-state driver it falls back to the default 48% speed (1584 MHz
in my case) quickly, and ramps up towards full speed equally
quickly.

 Testing builds of gcc-4.7.2 shows the new 'performance' setting is a
little faster (with -j4)  than the old acpi 'ondemand'.  A kernel
(x86_64) -j4 defconfig build is, however, marginally slower and one
build (hopefully just an outlier!) was quite a lot slower [ typically
6m52 for acpi, 6m57 for intel_pstate but that one build took 7m20 ].

 So, for kernels >= 3.9 on SandyBridge use the 'performance' scaling
governor instead of 'ondemand' if you enable this new driver [ e.g.
test for the /sys/devices/system/cpu/intel_pstate/ directory in your
own initscript, or perhaps change the default setting in your
.config ].

 This was a public information bulletin.

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce
--

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Gmane