console settings to be compatible with ubuntu
up to yesterday, i never had problems with the charsets, but now i finally could convince one of my customers to use linux on a new desktop machine. up to now, i just use linux for servers, mostly lfs and the desktops are running windows. most customers require one or another proprietary software which requires windows, and wine is usually not build for those special products. but that's not the question today, i just wanted to introduce the envrionment i'm useing. as mentionned, i never had any problems with this setup, but now i'm stuck with codepages!! i never had any problems to access the server from any windows. but with ubuntu i just get garbage with the filename and even with plain textfiles!! same with samba and nfs. i tried for hours to get a working configuration, but got no running result! as i'm from CH, i use KEYMAP="de_CH-latin1" FONT="lat9w-16 -m 8859-15" in /etc/sysconfig/console and dos charset = cp850 unix charset = ISO8859-1 in samba.conf first i played arround with the console settings (UNICODE= and LEGACY_CHARSET) but got no useable result. then i googled and got quite some hints, but none solves my problem. i built a fresh server with just samba and nfs to play arround. but creating a file like 'öä.txt' either on the server or on the ubuntu-workstation (no change from a 9.10 to 10.4 life system) the other system just shows up a garbage-filename. for every change i made to either samba or console i rebooted the server, as i'm not shure wether a 'console start' would really do the job. i really would like to have a first customer with a production-system on ubuntu, but if i can't fix the behaviour, i probably will have to install windows as any windows (from 98 to 7 runs fine with the above settings). thanks for any hint or help tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: perl, gdbm and perl's obsequious help
> And this may very well be the issue, since nobody has reported this > problem when building LFS before. not quite true! i had issues with perl, but everyone on this list told me i made a mistake. as i couldn't track the problem down to it's realy cause, i just gave up argueing. and the docs about how to build perl couldn't help, even google and company didn't help. the problem arises with some life-distors, but does not on others. as i was not able to see WHY, i just switched over to a distro which didn't screw up chapter 5. the perl built in chapter 5 compiles fine, but in chapter 6 as the libraries are not found any longer dies. i will add these lines to my scripts in the future and am glad someone really could track down the problem to its source. (and i am happy so see it really was not my fault). btw: i guess it's not only gdbm but other libraries too. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: 70-persistent-net.rules
Mike Johnston schrieb: > Hello,I have a LFS system with a read only file system. I have > /etc/udev/rules.d a symlink to a read/write partition. The > 70-persistent-net.rules file gets generated and keeps on growing for every > reboot. If i make the root filesystem read/write, this file does not get > re-generated and stays constant. > Any ideas as I need the root filesystem r/o and I/m imaging these systems in > our lab. > Mike look at /lib/udev/write_*_rules my solution is not very 'nice' but suites my needs: insert just an 'exit 0' after the '!/bin/sh -e' i had the problem with booting from an usb-stick which added all found network-cards and cd-drives to the persisten rules. i hope this helps tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: 70-persistent-net.rules
Mike Johnston schrieb: > Thank you. > That would work however, I want to make 30 instances for 30 machines. I need > to generate the 70-persistent-net.rules file so each machine has fixed > interface names. you don't need a 70-persistent-net.rules file. as you are writing in plural, you have more then 1 interface on each machine? if not, just forget about - the single interface will always be eth0. if you have more than 1, but with different chipsets, you can either create specific rules for udev or have 1 interface in the kernel (will be eth0) and the rest as modules, forcing the modules to be loaded in a given sequence resulting in the same order each time you boot. i'm having a board here with 2 onboard nics (realtec, but with abit mac-prefix 00-50-8d). the mac differs by 1 in the very last digit. booting without rules is 50% chance to have the nics the way you want. in this situation you really need to have the 70..rules-file! > What I don't understand is that if its a symlink on read/write partition it > consistently keeps adding to the file. If it's not a symlink but a real > file, it doesn't regenerate and stays a fixed size. > How does it know this? im not shure, but i guess it might be a feature (or bug) in the /lib/udev/rule_generator.functions i guess the directory is not writeable, just the linked file is. you can change the write_net_rules to not check for writeable or use another path for the locking-file by copying and renaming the fuctions from rule_generator.function to write_net_rules and modify it to your needs. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Bug in GNU make 3.82 ?
Dr.-Ing. Edgar Alwers schrieb: > Hi, > > building gstreamer-0.10.24 with make version 3.82 ends with an make error: > > "Makefile: 878: *** missing separator (did you mean TAB instead of 8 > spaces?). Stop " > make: *** [all] Error 2" make requires tabs, not spaces. this is documented quite a long time now. up to and including 3.81 spaces were accepted. thus 3.82 behaves as expected. not make has to be blamed, but the packages that don't accord to the documented specs. tobias -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
Re: make error with glibc on Fedora 14
> Rosario Turco virgilio.it> writes: remove the following blanks (^) and add the missing \ >> ../glibc-2.12.1/configure --prefix=/tools \ >> --host = $ LFS_TGT --build =$(../ glibc-2.12.1/scripts/config.guess) \ ^ ^ ^ ^ ^ >> --disable-profile --enable-add-ons \ >> --enable-kernel = 2.6.22.5 --with-headers=/tools /include \ ^ ^ ^ >> libc_cv_forced_unwind=yes libc_cv_c_cleanup= yes ^\ >> libc_cv_visibility_attribute=yes libc_cv_broken_visibility_attribute=no -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Firefix with GCC 4.6
firefox 3.6.16 fails on "make" with "nsEnumeratorUtils.cpp" with gcc 4.5.2 i had no problem. (i just testet my scripts sucessfully with gcc 4.5.2 and now restarted again with just replacing gcc 4.5.2 with 4.6.) i tried firefox 4.0 with no success it fails alreday during configure with "cant't find header GL/glx.h" but mesa is installed and /usr/X11/include/GL/glx.h exists and a symlink for /usr/include/GL to usr/X11/include/GL doesn't solve the problem. i did not try firefox 4.0 with gcc 4.5.2. i'd like to stick to the 3.6 for a while as not all plugins i use are available for 4.0. thus my favourite would be a fix for firefox 3.6.16 / gcc 4.6, but i could live with firefox 4.0 if somebody can give me a hint what i have to do to convice firefox to see the installed mesa... thanks for any help tobias -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
coreutils date - year 2038 bug
running the "certificate authority certificates" scripts from blfs my system fails with some certificates. as i learned this is not a coreutils bug but a kernel problem. is there a solution for 32bit kernels or do i have to rebuild with 64bit? thanks tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
xinetd
baldu...@units.it schrieb:
>
> are you sure that xinetd is picking libtirpc up?
>
> pmap_{set,unset} are in libtirpc and in order to force the xinetd build
> process to link in libtirpc I had to:
>
> export LDFLAGS="-ltirpc"
that's it. now libtirpc is linked.
but now i get the next errors:
libtirpc.so: undefined reference to
key_encryptsession_pk
getnetname
_des_crypt_call
getpublickey
key_gendes
applying the debinan-patches to remove the crypt-stuff, it's not really
getting better. '_des_crypt_call' is not mentionned any longer, but now
another 2 references are undefined:
cbc_crypt
ebc_crypt
meanwhile i went on without xinetd. with lsof i hit the same problem.
appending '-ltirpc' solves the 'pmap_*' issues as with xinetd and then
fails with the same undefined references from libtirps as xinetd.
thus i guess something's wrong with my libtirpc build.
i tried just cmi first, and then i applied the debian patches with the
above mentionned result.
i wonder which packages will fail too the next days... xinetd is not
vital (as mentionned by bruce). and lsof is just nice to have. but i
assume i'll find more packages and some might be a no-go...
thanks for your help
tobias
--
http://linuxfromscratch.org/mailman/listinfo/blfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
[lfs-support] 6.17 gcc-4.6.2 - problem with libmudflap tests
last time (3 days ago) i compiled gcc 4.6.2 on this machine (intel i7-2600 with 8gb ram) the compile run 2 minutes, the tests 16min. as far as i can see, i used the 7.0 release almost to the letter... (i upgraded the kernel and kernel-headers to 3.1.4) using this 7.0 system as base, i wanted to restart with current dev. compiling gcc 4.6.2 still needs 2 minutes, but the tests run for incredible 11h37mins. i had 429 failures in the libmudflab part (cat LOG | grep libmudflap.c|grep FAIL|wc -l) almost all of this failures are followed by WARNING: program timed out cat LOG|grep "timed out"|grep WARN|wc -l : 250 the statistics say 998 expected passes and 429 unexpected failures (same number as my grep/wc) what is exactly the same numbers as my last compile. the difference is the current run adds the 'program timed out' warnings and needs 11:26h more time to get the same result ;( a quick look in the log shows no other strange entries. my scripts are stil running, without any further problems. i just was puzzled when i came home, as i expected to have either the job done or a failure in a package. but gcc test running all day long i've never seen before! all binaries built with the current job (current dev) are slightly smaller then the files from the last run (7.0) as i couldn't find anything with google, i'm asking for help here. anybody any idea what i might have screwed??? the build already finished the LFS stuff and started with the packages i use from BLFS. i'll report tomorrow wether the system is stable or not. i guess it will be stable. i just would like to know WHY the libmudflap test are running for ever, with timeout on every single test. thanks for any help tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] gmp 5.0.5 + mpfr 3.1.0 failure
mpfr configure fails with:
...
checking for recent GMP... yes
checking for __gmpz_init in -lgmp... no
configure: error: libgmp not found or uses a different ABI (including
static vs shared).
Please read the INSTALL file -- see "In case of problem".
libgmp seems to be properly installed:
ld.so.cache points to the correct locations for libgmp and libgmpxx, ldd
on both has no missing dependencies, libgmp.la and libgmpxx.la both have
an empty old_library directive (thus i assume both are dynamic only as i
always configure with --disable-static).
neither the INSTALL nor google was of any help - at least none of them
seems to fit my configuration, most were about compiling on apples.
having a closer look at gmp, i had to see some failures during the tests:
...
PASS: t-fat
mpn_get_d wrong on 2^n
bit 0
exp 0
want_bit 0
sign 0
n=0x1
nsize1
want =[00 00 00 00 00 00 F0 3F] 1
got =[00 00 10 00 00 00 00 00] 5.1806537865363093631e-318
/bin/sh: line 5: 10259 Aborted ${dir}$tst
FAIL: t-get_d
...
PASS: t-bin
mpz_get_d wrong on 2**0
z=1
want 1
got 5.1806537865363093631e-318
/bin/sh: line 5: 15711 Aborted ${dir}$tst
FAIL: t-get_d
mpz_get_d_2exp wrong on 2**1
z=0x2
want =[00 00 00 00 00 00 E0 3F] 0.5
got =[00 00 10 00 00 00 00 00] 5.1806537865363093631e-318
want exp 2
got exp 2
/bin/sh: line 5: 15734 Aborted ${dir}$tst
FAIL: t-get_d_2exp
...
PASS: t-aorsmul
mpz_cmp_d wrong (from check_low_z_one)
got 1
want 0
x=2
y 5.18065e-318
x=0x2
y 5.18065e-318
y 00 00 10 00 00 00 00 00
/bin/sh: line 5: 16072 Aborted ${dir}$tst
FAIL: t-cmp_d
...
PASS: t-equal
mpq_get_d wrong on 2**-97
q=1/158456325028528675187087900672
want =[00 00 00 00 00 00 E0 39] 6.3108872417680944433e-30
got =[00 00 10 00 00 00 00 00] 5.1806537865363093631e-318
/bin/sh: line 5: 17208 Aborted ${dir}$tst
FAIL: t-get_d
...
PASS: t-fits
should be one ulp from 1: -inf
/bin/sh: line 5: 19520 Aborted ${dir}$tst
FAIL: t-get_d
mpf_get_d_2exp wrong on 2**-513
f=0x0.8@-128
want =[00 00 00 00 00 00 E0 3F] 0.5
got =[00 00 10 00 00 00 00 00] 5.1806537865363093631e-318
want exp -512
got exp -512
/bin/sh: line 5: 19543 Aborted ${dir}$tst
FAIL: t-get_d_2exp
...
PASS: t-locale
t-misc.cc:119: GNU MP assertion failed: d == 123.0
/bin/sh: line 5: 22158 Aborted ${dir}$tst
FAIL: t-misc
...
as the book reads the tests as critical, i guess the libgmp and/or
libgmpxx might really be broken.
googling about the 'get_d'-failures and the 'assertion failed' mentonied
problems with CFLAGS. as i don't set them, i have no clue where to dig on...
snippet from 'configure' for mpfr:
checking for CC and CFLAGS in gmp.h... yes CC=gcc -std=gnu99 CFLAGS=-m32
-O2 -pedantic -fomit-frame-pointer -mtune=pentiumpro -march=pentiumpro
checking for CC=gcc -std=gnu99 and CFLAGS=-m32 -O2 -pedantic
-fomit-frame-pointer -mtune=pentiumpro -march=pentiumpro... yes
checking for gcc... gcc -std=gnu99
checking whether the C compiler works... yes
same for gmp (ABI=32 ./configure...):
checking whether to enable maintainer-specific portions of Makefiles... no
checking ABI=32
checking compiler gcc -m32 -O2 -pedantic -fomit-frame-pointer ... yes
checking compiler gcc -m32 -O2 -pedantic -fomit-frame-pointer has
sizeof(long)==4... yes
checking compiler gcc -m32 -O2 -pedantic -fomit-frame-pointer
-mtune=pentiumpro... yes
checking compiler gcc -m32 -O2 -pedantic -fomit-frame-pointer
-mtune=pentiumpro -march=pentiumpro... yes
checking for gcc... gcc
checking whether the C compiler works... yes
as far as i understand, the relevant options m32, march and mtune are
the same.
thanks for any help
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
[lfs-support] gmp 5.0.5 + mpfr 3.1.0 failure
Bruce Dubbs schrieb: > Tobias Gasser wrote: >> mpfr configure fails with: >> >> >> ... >> checking for recent GMP... yes >> checking for __gmpz_init in -lgmp... no > > I'm not sure what's happening, but I have: > > $ nm /usr/lib/libgmp.so |grep gmpz_init > 00019a60 T __gmpz_init > 00019a90 T __gmpz_init2 > 0001a9c0 T __gmpz_init_set > 0001aa40 T __gmpz_init_set_d > 0001aa80 T __gmpz_init_set_si > 0001aaf0 T __gmpz_init_set_str > 0001ab50 T __gmpz_init_set_ui > 00019b10 T __gmpz_inits strange. > Perhaps a mistake in installing gmp? i did it as in the book... but i'll redo again. thanks for the hint! tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] gmp 5.0.5 + mpfr 3.1.0 failure - part. solved
Bruce Dubbs schrieb: >> checking for recent GMP... yes >> checking for __gmpz_init in -lgmp... no > i had a typo with the 'adjust'. thus ld was searching the wrong places... meanwhile the basic system is up and running tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] gmp 5.0.5
i still have the errors in the gmp tests. as the book says the tests are critical, i'd like to fix them. but as already mentionned, google was no help - at least not for me. is anybody else able to reproduce the errors? tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [blfs-support] GREP question
alex lupu schrieb: > As you can imagine, that leaves the PIPESTATUS array of the house in terrible > shape, at least Error 141, if not higher. to grin or not to grin is here the question... > I feel that THERE MUST BE a clean way to do it in bash where the main stream > can run to its natural completion unaffected by what goes on beyond (to the > right of) the TEE (the 'tee' command). no way at all. the first process sends all output to the pipe. the second process accepts input from the first thru the pipe. if the second process dies or finishes, the pipe will be closed. the first process can't deliver the output any longer, thus most programs will abort with an 141 code. some others will just hang around forever waiting for someone to listen. as mentionned in my first reply, you have to use temporary files - and take care not to run out of disk-space, or you'll have not the same but a comparable problem again... mit freundlichen grüssen tobias gasser -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
[lfs-support] glibc 5.7
why are the CC= AR= RANLIB= not set for glibc? as far i can understand, glibc will be built with the hosts compiler and not with the new one build in 5.5 as the new one didn't install reachable with the PATH-variable. in 5.8 binutils and 5.9 gcc will be built with the new compiler by specifying the CC/AR/RANLIB variables. further on these variables are no longer required, as in 5.9 the compiler will be installed to be found in the path. i just built glibc with CC/AR/RANLIB set as in 5.8 and 5.9. meanwhile i've rebuild glibc in chapter 6 without any issues. is there a good reason to build glibc with the host compiler instead of building with the new one from 5.5? tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] glibc 5.7
Andrew Benton schrieb: > On Thu, 21 Jun 2012 11:57:19 +0100 > Tobias Gasser wrote: > >> why are the CC= AR= RANLIB= not set for glibc? > > To quote from the glibc page > http://www.linuxfromscratch.org/lfs/view/development/chapter05/glibc.html > > "--host=$LFS_TGT, --build=$(../glibc-2.15/scripts/config.guess) > > The combined effect of these switches is that Glibc's build system > configures itself to cross-compile, using the cross-linker and > cross-compiler in /tools." > oops. sorry for the noise and thanks! tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] Q: Is the map between physical SATA disks and GRUB2's hd fixed?
Jeremy Henty schrieb: > Can I be sure that GRUB2's hd0, hd1 etc. will always correspond to the > same physical SATA connectors on the motherboard, no matter what > hardware I plug in? (I know from experience that /dev/sda does not > always map to the same connector.) If not, how can I find out the > mapping after booting from a live CD? Or can I drop into the BIOS and > asssume that it lists the drives in hd0, hd1 ... order? short: no, you can't. long: i'm using mobile racks in my dev systems. usually i just have 2, but in my 'big iron' i use a icy-dock 3disk cabinet (mb 973sp) which is connected to the 3 first sata-connectors. on the 4th i have the dvd. if i just have 1 disk inserted, it will always be hd0 and /dev/sda, independent which of the 3 slots the disk is in. if i insert 2 disks, the topmost (slot 1 or 2) will be hd0, and the lower one (slot 2 or 3) will be hd1. no matter how i populate (1+2, 1+3 or 2+3) the topmost is hd0/sda, the lower one hd1/sdb. inserting a 'next' disk while having lfs up will give the next letter, thus inserting a second will be /dev/sdb and the third /dev/sdc, independent of which slot i use. even worse #1: booting with 3 disks, i get hd0,hd1 and hd2 with grub, and /dev/sda, sdb and sdc. hd0/sda will be the topmost disk in the cabinet. unmounting and removing sdb and sdc, waiting some time and then reinserting sdc (the 3thd) will show it now as sdb, insering the former sdb a little later will show it as sdc! even worse #2: the mainboard has an additional marvel chip with another 2 sata ports. i have another 3.5" plus a 2.5" bay attached to this ports. if i just have inserted the 2.5" disk, grub assignes hd0, lfs uses /dev/sde. lfs will use sde and sdf for the 2 ports. having inserted just one it will be sde, independend wether i have the 2.5 or 3.5 bay filled on startup. i guess the onchip driver is loaded first (assigning sda-scd) and the marvel second (assigning sde+f). grub seems to enumerate all available ports just enumerating the found devices without leaving any 'whole'. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] Grub 2.0 fails on older intel atom board
i'm going to upgrade some machines with Asrock A330GC boards with Dual-Core Atom 330 processors. they all are running lfs dated may 2011, kernel 2.6.39, gcc 4.5.2. i built current dev without any problems, but booting was impossible. grub does not show anything. after the bios messages i just have an black screen with a blinking curser top left. waiting for 2 hours does not change anything... the disk does not start in any of 4 a330gc machines, but runs fine on some other machines with newer processors. (even on a asus at5nm10t with atom d525). after booting a ubuntu 10.04 lts live-cd i installed grub from ubuntu (saving the grub.cfg first and replacing the ubuntu autogenerated grub.cfg afterwards) and now the system runs fine as expected... (the system still has grub 2.0 installed, but i replace the bootloader and the /boot/grub stuff by just 'grub-install' from within ubuntu). restarting from scratch with grub 1.99 instead of 2.0 with everything else the same, the system boots as expected. i couldn't find any config options for grub to address this problem, and searching google and the grub bugtracker didn't help too. probably we are the first to install grub 2.0 on those older atlon boards... if somone else can confirm the failure with grub 2.0 on older atlons, i would suggest a warning in the grub page. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] remove or hide the build tools - was LFS 7.2 GCC pass 1
Am 03.09.2012 19:24, schrieb Israel Silberg:
> And another question, if I want in the end tohave a LFS or BLFS system
> without gcc in it, shold I keep it in the toolchain or should I remove it
> when the system is done?
i just add
--bindir=${DEVLOP}
--sbindir=${DEVLOP}
to the following packages configure-options:
gcc glibc automake autoconf pkgconfig libtool binutils
there might be some other packages to consider, but for me these seem to
be sufficient.
root and lfsuser (the user who builds all the stuff) have the following
added to their profile
DEVLOP=/home/lfsuser/devlop
PATH=${DEVLOP}:${PATH}
no user but root and lfsuser have access to ${DEVLOP} by "chown
lfsuser.root ${DEVLOP}"
everything but bin and sbin files are install in the usual place. but
the binaries are some kind of hidden.
binutils is a little special:
some binaries are installed in ${DEVLOP} AND /usr/bin (ar as ld ld.bfd
objcopy objdump ranlib strip) which i delete from /usr/bin to have juse
one copy in ${DEVLOP}
if there is need, you can just move a file from ${DEVLOP} to /usr/bin.
i do so for size and strings from binutils
and i had to make a symlink from ${DEVLOP}/cpp to /lib/cpp as some
packages require /lib/cpp (don't ask which - it's quite some time ago i
'invented' this 'security scheme' for my servers)
to be even more paranoid, just move all the stuff in ${DEVLOP} to an usb
stick - it's less than 8mb of data! whenever you have to build
something, just insert the stick and mount it to ${DEVLOP}. that's what
i did first. but as i have to maintain the servers over ssh (or travel
for hours), i dropped it - except for the client-systems at our local
school.
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] LFS 7.2 GCC pass 1
Baho Utot schrieb: > > I always build all of LFS with -j4 or -j8 and it has not failed me. > so you just had very very much luck on your way... there are packages which CAN fail with parallel builds. whenever i run into a problem, i just restart the package with -j1 which often solved the problem magically in the past... i'm used to set -jX to the number of cores i have (currently -j8 on my i7) globally which speeds up the build-time significally. my system might be a little special, as i use a SSD and the working diretory for the builds is a tmpfs as i have 16g of ram available. thus disk-io is very very very fast. from the base LFS i have to overwrite -j8 for make grub make groff make / test udev test tar test patch test binutils as i don't check the makeflag on each new version, probably one or another package meanwile would rund fine with -j8. for python make and test run fine with -j8, but install requires -j1 (not always, but on rary occasions it fails as files are tried to be installed into a directory not yet build by another thread) recently i had intermittent failures on both xulrunner and firefox where i build with -jX since a very long time. with version 15.0 the build fails at about half of the builds on my system. giving -j1 runs fine. i can use 8 cores in the mk_add_options MOZ_MAKE_FLAGS="-j8" but NOT in the MAKEFLAGS or the make-commandline! tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] check 0.9.9 (5.13) fails
version 0.9.8 compiles fine, but the new 0.9.9 fails with gcc -DHAVE_CONFIG_H -I. -I.. -I../src -I../src -g -O2 -Wall -ansi -pedantic -Wextra -Wstrict-prototypes -Wmissing-prototypes -Wwrite-strings -Wno-variadic-macros -MT check_thread_stress-check_thread_stress.o -MD -MP -MF .deps/check_thread_stress-check_thread_stress.Tpo -c -o check_thread_stress-check_thread_stress.o `test -f 'check_thread_stress.c' || echo './'`check_thread_stress.c mv -f .deps/check_thread_stress-check_thread_stress.Tpo .deps/check_thread_stress-check_thread_stress.Po /bin/sh ../libtool --tag=CC --mode=link gcc -g -O2 -Wall -ansi -pedantic -Wextra -Wstrict-prototypes -Wmissing-prototypes -Wwrite-strings -Wno-variadic-macros -o check_thread_stress check_thread_stress-check_thread_stress.o ../src/libcheck.la ../lib/libcompat.la -lrt libtool: link: gcc -g -O2 -Wall -ansi -pedantic -Wextra -Wstrict-prototypes -Wmissing-prototypes -Wwrite-strings -Wno-variadic-macros -o .libs/check_thread_stress check_thread_stress-check_thread_stress.o ../src/.libs/libcheck.so ../lib/.libs/libcompat.a -lrt -Wl,-rpath -Wl,/STAGE1/lib /STAGE1/lib/gcc/x86_64-unknown-linux-gnu/4.7.2/../../../../x86_64-unknown-linux-gnu/bin/ld: check_thread_stress-check_thread_stress.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5' /STAGE1/lib/gcc/x86_64-unknown-linux-gnu/4.7.2/../../../../x86_64-unknown-linux-gnu/bin/ld: note: 'pthread_create@@GLIBC_2.2.5' is defined in DSO /STAGE1/lib/libpthread.so.0 so try adding it to the linker command line /STAGE1/lib/libpthread.so.0: could not read symbols: Invalid operation collect2: error: ld returned 1 exit status the only differences to the book are the build-path (i use /STAGE1 for years without problems) and the kernel version: i use 3.4. as it is a long-term kernel. i don't understand why 'GLIBC_2.2.5' is referenced, as the host-system is lfs 7.2 with glibc 2.16.1 and the lib in /STAGE1/lib is glibc 2.16.1 too. check 0.9.8 compiles fine. any idea what's wrong? thanks tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] check 0.9.9 (5.13) fails
Am 12.11.2012 11:15, schrieb Tobias Gasser: > /STAGE1/lib/gcc/x86_64-unknown-linux-gnu/4.7.2/../../../../x86_64-unknown-linux-gnu/bin/ld: > note: 'pthread_create@@GLIBC_2.2.5' is defined in DSO > /STAGE1/lib/libpthread.so.0 so try adding it to the linker command line i was a little hasty writing my previous mail. adding CFLAGS="-L/STAGE1/lib -lpthread" make... fixes the problem configure confirms to miss libpthread, but does not throw an error! checking whether unsetenv is declared... yes checking for the pthreads library -lpthreads... no checking whether pthreads work without any flags... yes checking for joinable pthread attribute... PTHREAD_CREATE_JOINABLE checking if more special flags are required for pthreads... no checking for gawk... /usr/bin/gawk tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
[lfs-support] glibc 2.16.0 (5.7)
the book copies the rpc headers to the host system. to avoid changeing the host, i use the same sed as in chapter 6 sed -e 's##"rpc/types.h"#' \ -i sunrpc/rpc_clntout.c i guess this should be changed in the book tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] check 0.9.9 (5.13) fails
Am 12.11.2012 19:10, schrieb Bruce Dubbs: > > check is only built in Chapter 5 and you don't mention your host system. lfs/blfs about 3 weeks old >I've seen the error before in BLFS and figured it was a mismatch in > autotools and used something similar to your workaround above. i use a global DESTDIR=xy which works fine on most packages. but with some packages i have to specify 'make DESTDIR=xy install' as the global variable seems to be overwritten somewhere. after hours of searching i gave up to understand why this happens... with gcc i tried to find out why the french langauge files are built. i set LANGUAGE=de and LINGUAS="de de_CH de_DE en en_GB" global in the profile. most packages do as expected, some just ignore it and build everything (--enable-nls) or nothing (--disable-nls), but gcc is very special by buildint as expected the local/de but adds the french language at least back until version 4.4. (i tried with de,it and got de,it plus fr) tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] glibc 2.16.0 (5.7)
Am 12.11.2012 19:15, schrieb Bruce Dubbs:
>
> Possibly, but the book is really only adding some headers to the host
> system. The problem should only come up when using LFS-7.1 as a host.
ok. not really a problem.
except for the system requirements like {d,b}ash or {g,}awk there is no
other package where the host has to be modified. that's why i prefer the
sed instead of copying the headers.
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
[lfs-support] Grub 2.0 problem
since about 1 year i am using the same harddisk. as the procution system
is a little outdated now, i wanted to make a 'fresh' disk.
to boot from this disk i had to start an ubuntu livecd and install grub
from this cd. the systems (i have built both 32bit and 64bit) can boot
and run fine.
but i can't install my own compiled grub as a boot-loader!
/dev/sda1 = /boot
/dev/sda2 = / (for 32bit)
/dev/sda3 = / (for 64bit)
/dev/sda4 = extended
/dev/sda5 = SWAP
/dev/sda6 = DATA
booting the system with ubuntus grub works fine. running "grub install
/dev/sda" from a chroot (dev, proc, sys are mounted with --bind) says
everyting is fine but booting results in the grub console with:
GRUB loading.
Welcome to GRUB!
error: disk 'hd0,msdos1' not found.
Entering rescue mode...
'ls' just shows a newline - an empty list!
rebooting the live-cd, reinstalling grub, putting my grub.conf into
place - the system boots fine.
i built grub as in the book.
"grub-install /dev/sda" has no errors:
Installatoin finished. no error reported
grub from ubuntu can boot either partitions (32 or 64bit), so does grub
from parted magic.
i tried both grub versions i built (32/64) but none can boot, both just
enter the console as mentionned above.
grub.conf is very basic and works fine with grub from ubuntu and parted
magic.
** cut
set root='(hd0,1)'
set timeout=10
insmod ext2
menuentry "linux 32bit" {
linux /boot/kernel-3.4.18-t32 root=/dev/sda2
}
menuentry "linux 64bit" {
linux /boot/kernel-3.4.18-t64 root=/dev/sda3
}
** cut
ubuntu installs into
/boot/grub and /boot/grub/locale where the modules are in /boot/grub
lfs has an additional /boot/grub/i386-pc where the modules are. i
already copied all from /boot/grub/i386-pc to /boot/grub but no change.
the boot symlink exists as required for a boot partition.
google was no help - i probably don't know what to ask for...
i'm trying to get grub up now for more than a week and have no more
ideas what i could try.
any help welcome!
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
Am 13.11.2012 03:20, schrieb Bruce Dubbs:
what i missed in my original message:
/sda1 is ext2
/sda2 and /sda3 are ext3 (first attempt was with ext4, but as grub
didn't work i made backups, reformatted with ext3 and restored).
> If /boot is a separate partition, then the linux lines should look like:
>
> linux /kernel-3.4.18-t64 root=/dev/sda3 ro
>
> note the you don't specify /boot there. From the viewpoint of grub,
> there is no /boot directory.
there is, as i have a symlink.
i've removed the /boot and added the ro, but as expected no change.
grub does not find the disk.
> I suspect that you installed grub from ubuntu without /boot mounted as a
> separate partition.
from the live-cd i mounted /dev/sda2 (or sda3) to /mnt and then
/dev/sda1 to /mnt/boot.
grub-install --boot-directory=/mnt/boot /dev/sda
installs ubuntus grub with success - and the boot-loader starts my kernel(s)
from within on of my systems (32 or 64bit), i use
grub-install /dev/sda
( i even tried with --boot-direcory=/boot with /boot mounted, but as
expected no change as /boot is default for --boot-directory)
> One thing to do is to drop to the grub command line and do:
>
> grub> ls (hd0,1)
just empty.
ls => empty
ls (hd0,1) => error: disk 'hd0,1' not found
ls (hd0,msdos1) => error: disk 'hd0,msdos1' not found
hitting TAB after typing "ls (" does nothing. i guess the disk can't be
found at all.
with ubuntus grub i get what i expect: hd0,msdos{1,2,3}
the main problem is, the boot-loader does not see the disk!
> For the ubuntu instances, try:
i don't have any intalled ubuntu, i just use the livecd to get grub
installed.
> linux (hd0,2)/boot/kernel-3.4.18-t32 root=/dev/sda2
> linux (hd0,3)/boot/kernel-3.4.18-t64 root=/dev/sda3
no. there are just empty (hd0,x)/boot as /dev/sda1 will be mounted
later. the kernels are at (hd0,1)/ (or as i have a symlink) (hd0,1)/boot
> The trick is to know which version of the grub configuration file is
> being used. A simple 'grub install /dev/sda' will assume that it is
> using /boot/grub/grub.cfg from where /boot is located when the install
> is run.
i have grub.cfg in (hd0,1)/ (hd0,1)/grub and (hd0,1)/grub/i386-pc
no symlinks but copies.
as grub does not see the disk, the question where grub.cfg should be
found is not yet of interest ;) - first grub has to find the disk.
first thing to solve is to make the disk available to the boot-loader, i
don't want to see the grub-console rescue mode:
GRUB loading.
Welcome to GRUB!
error: disk 'hd0,msdos1' not found.
Entering rescue mode...
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
Am 13.11.2012 18:08, schrieb Bruce Dubbs:
> Tobias Gasser wrote:
>> Am 13.11.2012 03:20, schrieb Bruce Dubbs:
>>
>> what i missed in my original message:
>> /sda1 is ext2
>> /sda2 and /sda3 are ext3 (first attempt was with ext4, but as grub
>> didn't work i made backups, reformatted with ext3 and restored).
>
> insmod ext2 is supposed to be able to handle ext2/3/4.
i have "insmod ext2" in my grub conf:
** cut
set root='(hd0,1)'
set timeout=10
insmod ext2
menuentry "linux 32bit" {
linux /boot/kernel-3.4.18-t32 root=/dev/sda2
}
menuentry "linux 64bit" {
linux /boot/kernel-3.4.18-t64 root=/dev/sda3
}
** cut
but as grub does not see the disk at all, neither grub.cfg is processed
nor the ext2-module is loaded.
> What symlink? I don't know if grub understands symlinks, especially
> from one filesystem to another.
works fine. ubuntus grub has no problems. the 'old' grub 199 had no
problems with it too. and it's not to another filesystem, it's simply to
have /boot/grub available on the boot-partition:
from within /mnt/boot after "mount /dev/sda3 /mnt" and "mount /dev/sda1
/mnt/boot") i do "ln -sf . boot"
> I agree that there is a problem between the BIOS and GRUB. When you
> install the ubuntu system, what are the contents of grub.cfg. Also what
> is the output of 'mount'.
once again: i did NOT install ubuntu. i just use the ubuntu life-cd to
boot into the live environment. i don't install. i just open a terminal,
mount my /dev/sda3 to /mnt and /dev/sda1 to /mnt/boot and do
'install-grub --boot-directory=/mnt/boot /dev/sda'
thus i have nothing of ubuntu except the boot-loader. not even the
grub-xx tools are from ubuntu, they remain my own compiled binaries. NOW
the system can boot, but as soon as i reinstall grub with 'install-grub
/dev/sda' after booting, my compiled LFS-version of grub fails to see
any disk.
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
Am 13.11.2012 20:26, schrieb Bruce Dubbs: > OK, so I now understand that both sda2 and sda3 are lfs systems. sorry for not being clear... and sda1 is /boot for both systems > What I suggest doing is > > mount -v /dev/sda1$LFS/boot > mount -v --bind /dev $LFS/dev as mentionned, that is what i did. proc and sys are bind-mounted too > It would probably be best to also mount /sys, /proc, /dev/pts, and > /dev/shm as in Section 6.2. /dev/pts and /dev/shm i did not mount. i now just mounted them too. going into chroot now i remove the stuff from ubuntu with rm -r /boot/grub > Then run 'grub-install /dev/sda' as usual: Installation finished. no error reported > Make sure the kernel is in /boot and /boot/grub has grub.cfg and the > i386-pc directory with the modules. looks good. as usual. i copy my grub.cfg (the one inlined in my first message) into /boot, /boot/grub and /boot/grub/i386-pc > Then exit chroot, umount the file systems and reboot. > > GRUB should see your disk. as usual. GRUB boots into the rescue shell. ls => empty ls (hd0,1) => error: disk 'hd0,1' not found ls (hd0,msdos1) => error: disk 'hd0,msdos1' not found and once again, booting the ubuntu live-cd. mounting /dev/sda1 to /mnt grub-install --boot-directory=/mnt /dev/sda now i copy my same grub.cfg into /boot/grub rebooting grub runs as expected. but it's not the bootloader i compiled, its the bootloader and all the modules from ubuntu. i made 2 more experiment: instead of ubuntu i started parted magic and used its grub. i used the 32bit version and mounted my 32bit partition and the 64bit parted magic with my 64bit partition. both boot as expected. one mor thing to mention: i tried grub 1.99 (just applied the same sed as in the book for grub 2.0). almost the same result as with grub 2.0.: GRUB loading. Welcome to GRUB! error: disk not found. Entering rescue mode... grub 2.0 says: "error: disk 'hd0,msdos1' not found." where as grub 1.99 says: "error: disk not found." "ls" from the grub-rescue-console is empty "ls (hd0,1)" differs a little: grub 2.0 says: "error: disk 'hd0,msdos1' not found." where as grub 1.99 says: "error: disk not found." i can reproduce the problem on 3 different systems: - intel i7 - intel atom - intel core2duo i connected the disk to each sata-connector on each board. i even tried to connect the boot disk on the first sata and some more disks to other ports to see what ls tells me. always the same, the bootloader wont find any disk, cd or floppy (the core2duo has a floppy) where as the ubuntu-grub lists all connected disk (going to the shell and use "ls"). i'd say it can't be a bios problem. i don't see any reason to blame my toolchain. i build grub to the letter of the book. it seems i'm the only one having problems with building grub 2.0. thus i guess the problem must be somwhere in my setup. i have no other package with problems. x11 with xfce, firefox, thunderbird, qemu/kvm, libreoffice all run fine. the grub tools seem to be ok as far i can see, it is just the bootloader which fails to find any disk. i probably have to live without a self-compiled grub for a while... tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
Am 14.11.2012 00:26, schrieb Bruce Dubbs: > > I don't know what is going on. How is the disk partitioned? > > fdisk -l /dev/sda bash-4.2# fdisk -l Disk /dev/sda: 300.1 GB, 300069052416 bytes, 586072368 sectors Units = Sektoren of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disk identifier: 0x0008462b Gerät boot. AnfangEnde Blöcke Id System /dev/sda1 *2048 206847 102400 83 Linux /dev/sda2 206848309268471536 83 Linux /dev/sda330926848616468471536 83 Linux /dev/sda461646848 586072063 2622126085 Erweiterte /dev/sda561648896923688951536 83 Linux /dev/sda692370944 1230909431536 83 Linux /dev/sda7 123092992 1538129911536 83 Linux /dev/sda8 153815040 1845350391536 83 Linux /dev/sda9 184537088 2152570871536 83 Linux /dev/sda10 215259136 28079513532768000 83 Linux /dev/sda11 280797184 586072063 152637440 83 Linux > What is the output of > > ls -l /usr/lib/grub/i386-pc/*.img bash-4.2# ls -l *.img -rw-r--r-- 1 root root 512 14. Nov 01:43 boot.img -rw-r--r-- 1 root root 512 14. Nov 01:43 cdboot.img -rw-r--r-- 1 root root 512 14. Nov 01:43 diskboot.img -rw-r--r-- 1 root root 28856 14. Nov 01:43 kernel.img -rw-r--r-- 1 root root 1024 14. Nov 01:43 lnxboot.img -rw-r--r-- 1 root root 2848 14. Nov 01:43 lzma_decompress.img -rw-r--r-- 1 root root 1024 14. Nov 01:43 pxeboot.img > > grub is acting like the drivers for your hw are not properly embedded in > it's image on track 0. yes. but i have no idea why this happens. the 2 grubs from ubuntu and parted magic run fine. it's just my self-compiled version where i have the problem. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
> You should have /boot/grub/i386-pc/core.img.
this file exists.
> First, check that /usr/sbin/grub-install and
> /usr/share/grub/grub-mkconfig_lib exist. I suspect that is OK.
yes.
> Reading through grub-install, at line 336. we should have:
[...]
> Hopefully some of the above will help you to figure out what is going on.
thanks for the detailed description!
as it's 23.45 here, i'll have to go to bed. i'll check it tomorrow.
i never had the idea to step thru grub-install. i guess you're right
assuming i can find where the error occurs. i'll do it in parallel with
grub-install from parted magic to be able see where the differences
occur. i already have a second disk ready with /dev/sda{1,2,3} on a
second system.
i'll report the results as soon as i have either a success or at least
the details where grub-install fails.
thanks for your help!
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
> You should have /boot/grub/i386-pc/core.img.
yes. it's size is 26433
> grub-install is a script. Lets try to look at that to see if we can
> figure it out. Adding a few echo commands can confirm some of the
> settings. You can also try 'grub-install --verbose /dev/sda'. You may
> want to add --recheck. See below.
--verbose is unknow, but --debug is available
if --debug is given, in linne 378 (grub-install)
setup_verbose="--verbose"
is set. thus i guess --debug is fine
> First, check that /usr/sbin/grub-install and
> /usr/share/grub/grub-mkconfig_lib exist. I suspect that is OK.
yes
> Reading through grub-install, at line 336. we should have:
>
> source_dir=/usr/lib/grub/i386-pc
> target=i386-pc
yes
> The file /usr/lib/grub/i386-pc/modinfo.sh shoul have:
> #!/bin/sh
>
> grub_modinfo_target_cpu=i386
> grub_modinfo_platform=pc
yes
>
> A few lines later, it should set
>
> grub_setup=/usr/sbin/grub-bios-setup
yes
> This is a binary executable.
>
> Later, it sets device_map="/boot/grub/device.map"
yes
> It would be good to see if this file exists. Mine has the contents:
>
> (hd0) /dev/sda
>
> --recheck should recreate device.map.
nope.
the file does not exist.
--recheck alone fails,
--recheck /dev/sda is fine, but no device.map is built
find / | grep "device\.map"
does not find the file
older grub versions had something like 'grub-mkdevicemap' as far as i
remember. but 2.00 doesn't have it any more.
i copied from the ubuntu-version to /boot/grub, but i still just reach
the grub-rescue-console. so probably the missing device.map is not the
problem.
> The script then copies a lot of files to /boot/grub/{i386-pc,locale} and
> possibly (not for lfs) /boot/grub/{themes,fonts}.
yes
> It runs grub-probe. It should result in /dev/sda1. This could be
> where the problem is:
>
> $sudo /usr/sbin/grub-probe --device-map="" --target=device /boot
> /dev/sda1
no problem, the response is /dev/sda1 as expected
> It then creates /boot/grub/grubenv which for me is just a lot of # marks.
yes.
> Next, figure out what it things $fs_module, $disk_module,
> $partmap_module, and $devabstraction_module should be by checking what
> $modules is.
fs: ext2
disk: biosdisk
part: part_msdos
modules: biosdisk ext2 part_msdos
looks ok for me
> It then runs grub-mkimage. See what the parameters are being used for
> that. This is what should create core.img.
i just put "echo" before "$grub_mkimage..."
in lines 720 and 722 and got:
bash-4.2# ./grub-install /dev/sda
/usr/bin/grub-mkimage -d /usr/lib/grub/i386-pc -O i386-pc
--output=/boot/grub/i386-pc/core.img --prefix=(,msdos1)/grub biosdisk
ext2 part_msdos
what is NOT what i expect
1) there is no load.cfg, but maybe that's ok
2) the prefix is WRONG, it should be either "/boot" or "(hd0,msdos1)"
i went back to line 642 an got
drive: (hostdisk//dev/sda,msdos1)
and after the sed
partition: ,msdos1
drive: hostdisk//dev/sda
i now just added
prefix_drive="(hd0,msdos1)"
in line 717 to force the correct value
bash-4.2# ./grub-install /dev/sda
/usr/bin/grub-mkimage -d /usr/lib/grub/i386-pc -O i386-pc
--output=/boot/grub/i386-pc/core.img --prefix=(hd0,msdos1)/grub biosdisk
ext2 part_msdos
which looks more reasonable to me.
> Finally, it runs grub-setup. See what parameters are being used for
> that also.
bash-4.2# ./grub-install /dev/sda
/usr/sbin/grub-bios-setup --directory=/boot/grub/i386-pc --device-map=
/dev/sda
(same with and without my inserted line 717)
i now run "grub-install /dev/sda" with my inserted line 717.
i'll be back in either 2 minutes after sucessful reboot, or 15 minutes
id i have to use the live-cd to reinstall grub...
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] Grub 2.0 problem
Am 16.11.2012 12:06, schrieb Tobias Gasser: > i now run "grub-install /dev/sda" with my inserted line 717. > i'll be back in either 2 minutes after sucessful reboot, or 15 minutes > id i have to use the live-cd to reinstall grub... still the same. i just get the grub-rescue-console. GRUB loading. Welcome to GRUB! error: disk 'hd0,msdos1' not found. Entering rescue mode... tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] A startup quesion
> Reading back, my sentence could be misleading, so, to be clear, I was > referring LO to build size, not install size (Bruce just pointed out > this, but I feel it needed to be close to my own statement). For install > size, it is about the same as OpenJDK, over 440MB, build size also the > same order for LO and OJDK. i need a little less for my 3.6.3.2 the download dir is little above 500m (since 3.6.0, no automatic delete available, impossible to remove manually due to the numeric prefixes - i guess i'll try to make a script one day...) the unpacked translations are about 1.3 (older versions had plenty of stuff in 2 more directories, 3.6.3 doesnt download or extract them any more) the working directory expands up to 3.7g install-size is just 350m (stripped!, only EN, DE, FR languages) max disk usage during build is about 5.8g (64bit, 5.5g with 32bit) i have an intel i7 with 16g ram and 320 wd raptor disk. last 32bit build used 3.4g build, 350m destdir and run for 50:48 last 64bit build used 3.7g build, 380m destdir and run for 52:35 tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] A startup quesion
Am 21.11.2012 14:38, schrieb Fernando de Oliveira:
> I thought in stripping it, but I am still not sure if any problem would
> appear. What is the command you used to strip?
i use DESTDIR on all packages, so i can loop thru all files in DESTDIR
to (re)compress all manpages, strip all binaries and libraries, set
chmod 755 on libraries and some more things...
here the snipet for stripping:
OK=$( LC_ALL=C file "${line}" )
if [ ! -z "$( echo ${OK} | grep ' ELF ' )" ] ; then
strip -p --strip-unneeded "$line"
elif [ ! -z "$( echo ${OK} | grep ' current ar archive' )" ] ; then
strip -p --strip-debug "$line"
fi
> Does this include the sources (compressed and uncompressed) as in my
> case? If so, please, would you post the switches used?
what else if not the sources???
i extract the libreoffice-core-3.6.3.2.tar.xz
then i make a symlink for the src as i don't want to download these 500g
on each build. if a new version requires new packages, the download
script will just get the new files. as the translations are extracted
into the src directory (another 1.3g!) i keep them to speed up the
build. as soon i have successfully built the new one once, i remove the
old translations and clean up the src
after 'make' i remove the src again, and i have 3.4g on 32bit or 3.7g on
64bit.
you can add the 1.3g translations and even the 500m src to my 3.7g
workdir giving a total of 5.5g. still far away from the 7g you mentioned.
here my autogen switches:
./autogen.sh \
--prefix=/usr/X11 \
--sysconfdir=/etc/libreoffice \
--disable-binfilter \
--disable-mozilla \
--disable-odk \
--disable-postgresql-sdbc \
--disable-kde \
--disable-kde4 \
--disable-gtk3 \
--disable-systray \
--enable-librsvg=system \
--enable-dbus \
--enable-extra-font \
--enable-release-build \
--with-system-boost \
--with-system-cairo \
--with-system-curl \
--with-system-db \
--with-system-expat \
--with-system-gettext \
--with-system-icu \
--with-system-jpeg \
--with-system-libpng \
--with-system-libxml \
--with-system-neon \
--with-system-nss \
--with-system-openssl \
--with-system-poppler \
--with-system-redland \
--with-system-zlib \
--with-system-mysql \
--with-system-lcms2 \
--with-system-mozilla \
--with-system-mesa-headers \
--with-num-cpus=$(getconf _NPROCESSORS_ONLN) \
--with-lang="de fr" \
--with-perl-home=/usr \
--with-openldap \
--without-java \
--without-system-jars \
--without-ppds \
--without-afms \
--without-myspell-dicts \
--without-system-dicts \
--without-help \
--without-helppack-integration \
--disable-unix-qstart-libpng
i don't build any dictionaries. i have most (but not all) libraries
already installed. disabling odk and binfilter saves some space (and
time) too.
as i'm running xfce, i need no kde or gtk3, just gtk2.
it's a pitty lo is not able to check the installed libraries. either you
use --with-system-xx (or as for librsvg --enable-xx=system) or lo will
build it's own package again! even better to check for installed
libraries first to avoid even the download - last time i counted, i
could save 30 out of 100 (as far as i remember nearly 200 out of 500g).
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
[lfs-support] coreutils
version 8.20 is available since oct 23. lfs-book has a ticket for it: 3215 searching the lists just show the ticket, but no further comments so far. the ticket says 'no announce yet', here it is: http://lists.gnu.org/archive/html/coreutils-announce/2012-10/msg0.html + bugfix release + speed improvement - nothing i can see what am i missing? thanks tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] can't compile gcc pass 1
My error with this page, even after having built one or two versions of
LFS, was that the last line:
"gcc compilation OK"
that made me ignore the other lines, when some of these lines were
telling me that I had requirements to fix.
I thought: "all right, host can compile, so the other lines are just
additional information or recommendations, may be".
i had the same problem. so i reworked the script.
i moved the compile test after checking for gcc, and put linux at the
end, as there have to be checked 2 versions
the output does not only show min requirement, but the current version
the book builds too.
my script aborts if /bin/sh is not bash, awk not gawk or yacc not bison.
if awk or yacc are scripts, i show a message to check the script.
tobias
#!/bin/bash
PATH=/usr/bin:/bin:/usr/sbin:/sbin
if [ -d /root/devlop ]
then
PATH=/root/devlop:$PATH
fi
# Simple script to list version numbers of critical development tools
export LC_ALL=C
echo "bash >= 3.2 (4.2)
$( bash --version | head -n1 | cut -d" " -f2-4 )
/bin/sh MUST be a link to bash"
TMP=$( readlink -f /bin/sh )
[ ${TMP//bash} == ${TMP} ] && echo "!! ERROR : ${TMP}" && exit 99
echo "/bin/sh -> ${TMP}"
echo "binutils (ld) >= 2.17 (2.31.1)
$( ld --version | head -n1 )"
echo "bison >= 2.3 (2.7)
$( bison --version | head -n1 )"
TMP=
[ -e /usr/bin/yacc ] && TMP=/usr/bin/yacc
[ -e /bin/yacc ] && TMP=/bin/yacc
[ -z "${TMP}" ] && echo "!! ERROR : yacc not found" && exit 99
if [ -L ${TMP} ] ; then
echo "yacc should be a link to bison"
TMP2=$( readlink -f ${TMP} )
[ ${TMP2//bison} == ${TMP2} ] && echo "!! ERROR : ${TMP2}" && exit 99
echo "${TMP} -> ${TMP2}"
else
echo "
!! IMPORTANT !!
check wether ${TMP} is a wapper script for bison
"
fi
echo "bzip2 >= 1.0.4 (1.0.6)
$( bzip2 --version 2>&1 < /dev/null | head -n1 | cut -d" " -f1,6- )"
echo "coreutils (chown) >= 6.9 (8.20)
$( chown --version | head -n1 )"
echo "diffutils >= 2.8.1 (3.2)
$( diff --version | head -n1 )"
echo "findutils >= 4.2.31 (4.4.2)
$( find --version | head -n1 )"
echo "gawk >= 3.1.5 (4.0.2)
$( gawk --version | head -n1 )"
TMP=
[ -e /usr/bin/awk ] && TMP=/usr/bin/awk
[ -e /bin/awk ] && TMP=/bin/awk
[ -z "${TMP}" ] && echo "!! ERROR : awk not found" && exit 99
if [ -L ${TMP} ] ; then
echo "awk MUST be a link to gawk"
TMP2=$( readlink -f ${TMP} )
[ ${TMP2//gawk} == ${TMP2} ] && echo "${TMP2} !! FAILURE !!" && exit 99
echo "${TMP} -> ${TMP2}"
else
echo "
!! IMPORTANT !!
check wether ${TMP} is a wapper script for gawk (GNU awk)
"
fi
echo "gcc >= 4.1.2 (4.7.2)
$( gcc --version | head -n1 )"
echo "gcc compilation"
echo 'main(){}' > dummy.c
gcc -o dummy dummy.c >/dev/null
TMP=
[ -x dummy ] && TMP=ok
rm -f dummy dummy.c
[ -z "${TMP}" ] && echo "!! ERROR : gcc compilation failed" && exit 99
echo "Compilation OK"
echo "glibc (ldd) >= 2.5.1 (2.17)
$( ldd --version | head -n1 )"
echo "grep >= 2.5.1a (2.14)
$( grep --version | head -n1 )"
echo "gzip >= 1.3.12 (1.5)
$( gzip --version | head -n1 )"
echo "m4 >= 1.4.10 (1.4.16)
$( m4 --version | head -n1 )"
echo "make >= 3.81 (3.82)
$( make --version | head -n1 )"
echo "patch >= 2.5.4 (2.7.1)
$( patch --version | head -n1 )"
echo "Perl >= 5.8.8 (5.16.2)
$( perl -V:version )"
echo "sed >= 4.1.5 (4.2.2)
$( sed --version | head -n1 )"
echo "tar >= 1.18 (1.26)
$( tar --version | head -n1 )"
echo "texinfo (makeinfo) >= 4.9 (4.13a)
$( makeinfo --version | head -n1 )"
echo "xz >= 5.0.0 (5.0.4)
$( xz --version | head -n1 )"
echo "linux kernel >= 2.6.25 (3.7.1), compiled with gcc >= 4.1.2 (4.7.2)
$( cat /proc/version )"
bash >= 3.2 (4.2)
bash, version 4.2.39(2)-release
/bin/sh MUST be a link to bash
/bin/sh -> /bin/bash
binutils (ld) >= 2.17 (2.31.1)
GNU ld (GNU Binutils) 2.23.1
bison >= 2.3 (2.7)
bison (GNU Bison) 2.7
!! IMPORTANT !!
check wether /usr/bin/yacc is a wapper script for bison
bzip2 >= 1.0.4 (1.0.6)
bzip2, Version 1.0.6, 6-Sept-2010.
coreutils (chown) >= 6.9 (8.20)
chown (GNU coreutils) 8.19
diffutils >= 2.8.1 (3.2)
diff (GNU diffutils) 3.2
findutils >= 4.2.31 (4.4.2)
find (GNU findutils) 4.4.2
gawk >= 3.1.5 (4.0.2)
GNU Awk 4.0.1
awk MUST be a link to gawk
/usr/bin/awk -> /usr/bin/gawk
gcc >= 4.1.2 (4.7.2)
gcc (GCC) 4.7.2
gcc compilation
Compilation OK
glibc (ldd) >= 2.5.1 (2.17)
ldd (GNU libc) 2.16
grep >= 2.5.1a (2.14)
grep (GNU grep) 2.14
gzip >= 1.3.12 (1.5)
gzip 1.5
m4 >= 1.4.10 (1.4.16)
m4 (GNU M4) 1.4.16
make >= 3.81 (3.82)
GNU Make 3.82
patch >= 2.5.4 (2.7.1)
GNU patch 2.7.1
Perl >= 5.8.8 (5.16.2)
version='5.16.2';
sed >= 4.1.5 (4.2.2)
GNU sed version 4.2.1
tar >= 1.18 (1.26)
tar (GNU tar) 1.26
texinfo (makeinfo) >= 4.9 (4.13a)
makeinfo (GNU texinfo) 4.13
xz >= 5.0.0 (5.0.4)
xz (XZ Utils) 5.0.4
linux kernel >= 2.6.25 (3.7.1), compiled with gcc >= 4.1.2 (4.7.2)
Linux version 3
Re: [lfs-support] can't compile gcc pass 1
Am 05.01.2013 18:39, schrieb Bruce Dubbs:
> The idea of the script was that it should be short. Generally the
> problem is that the symlinks are not set and occasionally makeinfo is
> not installed. Rarely is the problem an out-of-date executable.
agree.
so why not check just the very important stuff?
check the versions for
- bash
- binutils
- gcc
- texinfo
i guess any distro matching at requested versions for these packages
should have the other packages requrements too.
more important than more package-versions are imho a working compiler
- gcc compilation
the kernel
- version
where as 'compiled with' should be met by the gcc version too. i am not
aware of any distro shipping the kernel compiled with a older version of
gcc than the kernel (ok, maybe the .point release does not match).
and the links
- sh
- yacc
- awk
as some distros seem to get rid of the /usr/{bin,sbin} my script checks
for /usr and /usr/bin. if one of the links fails, i STOP the script with
an error.
i guess with an approach like that, less novices will fail in the very
beginning.
just my thoughts...
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] can't compile gcc pass 1
> AWK=`readlink -f /usr/bin/awk`
> awk=/usr/bin/awk
> [ "$AWK" == "/usr/bin/gawk" ] || die "$awk is not a symlink to gawk"
some distros started to drop the /usr hierarchy
the script should be a little smarter to accept the files not only in
/usr/bin but just anywhere in $PATH
something like
awk=$( which awk )
AWK=$( readlink -f "${awk}" )
[ "${AWK%%\/gawk}" == "${AWK}" ] && die...
tobias
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
Re: [lfs-support] gperf in glibc-2.17
Am 22.02.2013 22:50, schrieb Frans de Boer: > I understand that gperf is not required - it would be mentioned > otherwise does it? yes. gperf is NOT required for glibc (or any other package in lfs) and yes, all requirements are mentionned. > However, I am not the first nor probably the last running into this > issue. Even glibc-2.16 requires the use of gperf and thus failed the i never had to build gperf for lfs. and i'm building lfs for quite some years (my first lfs was 3.0). glibc builds fine without gperf if you follow the book. i think you are the first with this issue. as far i can remenber i never have read about gperf issues so far. but as i'm getting older, my memory seems to have failures sometimes ;) i just can say/write: FBBG (follow book, book good!) if you have issues about missing gperf in lfs, you probably dont follow the book. if you explain where you deviate from the book, maybe someone can help to find the 'wrong' part. btw: later in blfs i build gperf as xcb-util requires it. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] gperf in glibc-2.17
Am 22.02.2013 23:29, schrieb Frans de Boer: > > Just search for "gperf glibc lfs". It is even on the archives by LFS. > thanks. i've got it. so you really are not the first... but the first part of my last message is stil valid: i never had to build gperf in lfs. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] gperf in glibc-2.17
Am 22.02.2013 23:56, schrieb Bruce Dubbs: > Looking at the glibc source, gperf only appears to arise from locale > issues. It would appear that you did not set up the lfs user in > accordance with the book: > > LC_ALL=POSIX > > No other LC_* or LANG settings should be set. > yep! reading your reply i rember having issues with my very first builds with LANG=de_CH in chapter 5. but i really don't remember which packages and what errors i had... unset LANG LC_ALL=POSIX is required for chapter 5 in chapter 6 i'm building with LANG=de_CH.UTF-8 LC_ALL=C btw: do you have problems with glibc in chapter 5 or 6? tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] lfs-support Digest, Vol 2802, Issue 1
Am 26.02.2013 15:24, schrieb Rick Berube: >> > > Just as a guess, I moved gmp to the Real Machine and re-attempted the > process. This time it was successful. I would infer that LFS doesn't > play well on virtualized hosts. > > Thanks. > i use qemu since about 6 months. before i used virtualbox (first from ubuntu, later i installed it in a blfs environment). i never had any problems to build lfs inside qemu or virtualbox. can you give more details where you guess to have problems with virtualisation? tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: [lfs-support] First LFS login/boot
> > I'm sorry Mr. Bruce, I missed it. > here is the result > - > root:/sys/class/net# dmesg|grep eth0 > [2.611331] via-rhine :00:12.0: eth0: VIA Rhine II at 0xfdffe000, > 00:e0:4d:56:48:0a, IRQ 23 > [2.612057] via-rhine :00:12.0: eth0: MII PHY found at address 1, > status 0x7849 advertising 01e1 Link > [ 20.191794] ADDRCONF(NETDEV_UP): eth0: link is not ready > [ 24.662664] via-rhine :00:12.0: eth0: link down > [ 24.663638] ADDRCONF(NETDEV_UP): eth0: link is not ready > [ 24.669753] ADDRCONF(NETDEV_UP): eth0: link is not ready as far i understand the above lines, you have sucessfully loaded the kernel driver for your device. the "link is not ready" probably means you have no ethernet cable connected, or the cable is broken. > My main purpose is creating Live CD with some unicodes support, should I > consider something about USB modem? no. usb-modems are obsolete - at least here in CH and DE. most users have adsl- or cable-routers to connect to the internet. just connect an ethernet cable (or use wifi) and use dhcp. the last usb-modems i got hands on was about 4 years ago. -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: /dev/null permissions after reboot
Manuel Gonzalez Montoya schrieb: > > root [ ~ ]# ls -l /dev/null > crw-rw 1 root root 1, 3 2008-03-24 12:44 /dev/null > something's wrong with your udev installation: in /etc/udev/rules.d/25-lfs.rules you have an line with KERNEL=="null", MODE="0666" 1) you've missed to install the rules.d files 2) you overwrite this in another rules-file -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: svn-book - Fdisk issue
David Kredba schrieb: > I used the current SVN version of the LFS book. > > The util-linux-ng version is 2.13.1. same version here > I am using UNICODE="1" in /etc/sysconfig/console and LANG is set > to cs_CZ.UTF-8. LC_ALL=de_DE.UTF-8 LANGUAGE=de LANG *unset* /etc/sysconfig/console UNICODE=yes KEYMAP="de_CH-latin1" KEYMAP_CORRECTIONS="euro2" FONT="lat9w-16" LEGACY_CHARSET="iso-8859-15" > Do you someone with a non "C" locale and UNICODE=1 have the same problem > please? works fine here. i have some issues with "special" chracters. fdisk w/o any arguments prints the usage screen with "..." in the last line (LC_ALL=C). having LC_ALL=de_DE.UTF-8 i get a fancy questionmark (which is the utf8 character 0x2026). try to set the LEGACY_CHARSET and FONT variables. i had very strange behaviour without them - now i just have strage characters on the screen. as recent cups requires UTF8 to export the printers to samba, i startet building a fresh LFS with utf8 just a week before as my current system looks very strange just having a few packages updated with utf8. i have actually no problems so far (but i'm not finished yet). when everything runs fine again i'll have a closer look at these special characters. in the meantime i'm living with them... -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: I really need help kernel panic
Spahn, Daniel schrieb: > Correction! > I read a little more carefully. You have not compiled the correct > drivers for your SCSI hard drive into the kernel. You can run lspci, > then make menuconfig (In a different terminal) and match up the SCSI > drive characteristics with a driver, most likely in the SCSI section of > Device Drivers. > This line here: > [ 13.472000] 0300 4194302 hda driver : ide-cdrom i guess it's not really scsi but sata, thus sata and the intel driver must be enabled in the kernel. built-in <*>, not modlue i suggest to use the "intel pata" for your cd/dvd, not the old ata thus you'll get no more /dev/hdX but you'll find your cd/dvd as /dev/sr0 with a link on /dev/cdrom, and your disk on /dev/sda with kernel 2.6.26.1 (the one i have access right now): device drivers SCSI device support <*> SCSI Disk <*> SCSI CDROM [ ] SCSI low-level drivers < > ata/atapi/mfm/rll <*> serial ata (proc) and parallel ata [*] ata acpi support [*] ata sff support <*> intel esb, ich, piix3... <*> intel pata mpiix -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Failure to boot
> > swapon: cannot stat dev/sda3: no such file or directory > fsck.ext3: No such file or directory while trying to open /dev/sda4 > > I suspect this is due to my grub configuration. Rather than installing > grub, I added these lines to menu.lst on my host system: > > title LFS 6.3 > root (hd0,3) > kernel /boot/lfskernel-2.6.22.5 root=/dev/hda4 > > Here is what my harddisk looks like: > > sda1 = NTFS > sda2 = Ubuntu > sda3 = swap > sda4 = LFS > i guess your kernel is missing the drivers for you disk controller. ubuntu has no /dev/hda but /dev/sda. for me this indicates your system has sata disks just have a look at the thread "i really need help kernel panic", you probably too didnt chose the correct sata modlues for your system. from within ubuntu try "lspci" on the console and check what controller your system has and activate the correct sata-driver. finally you'll have to change /dev/hda4 to /dev/sda4 in your menu.lst i hope this will solve your problem tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Failure to boot
> > 00:1f.1 IDE interface: Intel Corporation 82801EB/ER (ICH5/ICH5R) IDE > Controller (rev 02) > 00:1f.2 IDE interface: Intel Corporation 82801EB (ICH5) SATA Controller > (rev 02) > > I assume the first device is my DVD/CDRW, so the other must be my > harddisk. Based on this, should I build the "Intel ESB, ICH... [etc] > support" module into my kernel and rebuild it? exactly. to access your cd/dvd activate "intel pata mpiix" too. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Failure to boot
> Root-FFS: No NFS server available, giving up > VFS: Unable to mount root fs via NFS, trying floppy > VFS: Cannot open root device "sda4" or unknown-block(2,0) > Please append a correct "root=" boot option; here are the available > partitions > 0300 156290904 hda driver: ide-disk > 0301 102398278 hda1 > 0302 29294527 hda2 > 0303 979965 hda3 > 0304 23615550 hda4 > 1600 4194302 hdc driver: ide-cdrom > Kernel panic -not syncing: VFS: Unable to mount root fs on > unknown-block(2,0) you get a listing of available partitions on HDA, found by the ide-disk driver. > For the sake of completeness, my grub entry is: > > title LFS 6.3 > root (hd0,3) > kernel /boot/lfskernel-2.6.22.5 root=/dev/sda4 as your log shows partitions found on HDA not SDA you probably have to change your grub menu to "... root=/dev/hda4" > > I just double checked with gparted to make sure lfs is indeed on sda4, > and it is. I also checked my kernel configuration and I did indeed > rebuild it with the proper modules. I also double checked the > timestamps of the bzImage in the source tree and the kernel image in > /boot to make sure that I didn't forget to copy it after it was > compiled. I'm clearly overlooking something, but I'm not sure what. the partitions found by ubuntu on SDAx will be found from your actual lfs-kernel according to the log on HDAx. you probably have still some old PATA drivers active. this driver seems to feel responsible for your disk on /dev/hda, and thus no driver is responsible for the same disk on /dev/sda. * second try: you copied the kernel to /dev/sda4/boot (on the LFS-partition) or /dev/sda2/boot (ubuntu)? copy it to both and play arround with root (hd0,X) kernel (hd0,Y)/boot/... where X/Y means all possible permutations with 1 and 3. (eventually try again with root=/dev/sda4 instead of hda4) tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Failure to boot
> Changing sda->hda in menu.lst an fstab allows me to boot, though it > still displays hda. The proper sata driver is built into the kernel. > From your post, it sounds like I probably have an older drive that's > taking precedence over the sata, which is causing this problem? > > Just for my information, boot when the disk is detected as hda rather > than sda isn't going to do any damage, correct? It would just read at a > lower speed? i don't think you'll have any performance issues with the pata driver. you can use "hdparm -tT /dev/xx" to test the read-performance (bufferd + cached). hdparm is part of blfs. as within then next major kernel versions the old pata drivers might get obsolete, all newer mainbords support sata, sata cabeling is blocking the air-flow inside the case much less and actually sata drives are already cheaper than pata i myself build the kernel without the old ata/atapi drivers now for quite some time. i just disable "device drivers" / "ata/atapi/mfm/rll" as i had the same strange behaviour on some mainboards when i enabled the ata/atapi to support an pata-cdrom. on one of the mainboards the disk changed from sda to hda when i changed the sata settings in the bios from legacy to enhanced (or something like this), on another it remained always hda as long as the general ide-driver was enabled in the kernel. (having sata in the kernel and ata/atapi as a module might solve the problem, but i didn't test it...) tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Grub: Error 21, Selected disk does not exist
Michael schrieb: > I'm going through Linux From Scratch v6.4, and I'm on Chapter 8.4. > Currently, I'm running Linux completely off of an external hard drive, while > I'm also building my Linux From Scatch build onto a seperate partition on > that disk. The laptop already has one hard drive in it, but I'm not really > touching it. > > Here's what happens: > > grub> root (hd0,4) > > Error 21: Selected disk does not exist > > As I said, we're on the external hard drive, sda. I'm not really sure what > sda2 is, so I've just sorta been ignoring it. sda5 is the LFS build, while > sda6 is the host system. I'm completely ignoring sdb1. your kernel seems to see the usb-disk first, the builtin disk second. thus after booting you have sdA external and sdB internal. don't assume /dev/sda=(hd0) and /sev/sdb=(hd1)!! simple trick to check how grub maps your drives: in the grub console type "root (hd0," and now the TAB key. grub shows all partitions on hd0. now repeat with "root (hd1,"+TAB to see hd1. compare the output with what you get from "fdisk -l". i guess, you'll see you have to use "root (hd1,4)" to specify your external disk. probably your current kernel has builtin support for IDE/ATA only, loading SATA drivers as modules, but AFTER usb support. thus your external usb disk is recognised first, the SATA second. i'd try to fix that first! it's more conveniant to have builtin before external devices. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
A Suggestion For A Simple Package Manager
> A good package manager is, IMO, a necessity. NO i'll explain why i NEVER will use any package-manager: have a look at the configure options with php5: depending on what other packages i've alread installed, the php package will contain the mcrypt/mhash ONLY IF NEEDED, same with mysql or pgsql, the imap-functionnality, the soap (xml) and quite some more. on a full blown server i have tonns of --with-xy or --enable-ab, after building the required libraries and tools. on a router where i have php available for my web-interface to maintain the system i just don't need them (and thus have lots of --without-xy and --disable-ab). which version do you declare the one and only to be used for the package-management?? with (b)lfs i have full control on what is installed on a system. i really dislike packages like "php5" + "php-mysql" + "php-pgsql" + "php-whatever-you-wish" (and each one again with -devel). i just compile php with all the options i need on a particular machine. or have a look at samba: for a small network i don't use ldap. but sometimes i use it and thus have to compile with --enable-ldapsam. same with cups or pam. usually i don't install swat (thus --disable-swat), but some customers are already used to it, thus i include it. greetings from switzerland tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: A Suggestion For A Simple Package Manager
> I'm a little surprised at the vehemence of your statement. > Why are you so excited by my opinion which applies only to > the machines I own that you used four exclamation marks? sorry, i didn't want to offend anybody. > ISTM that you have relatively little contact with > reasonable version control, package management, and > build control systems. It shouldn't even be necessary i think i have a basic knowlege about it, but i don't use them. > to compile or build the image(s) necessary for a given > machine on a machine of the same architecture, let > alone for it to be the same machine. Most of the work again i have theoretic knowledge about cross-compiling. an i often build a lfs distro on a powerfull machine to transfer it at the end to an other machine (but always in the same family, i.e. without the need of cross-compiling). [...] > Of course, creation of such a collection of tools is a > significant effort, as is maintenance of it, as well > as the maintenance of the specifications, though those > should be relatively static by comparison with the > point releases of the components comprising the deliverable. for me, the effort to maintain this tools would be to big. i'm now used to maintain my batch-scripts to have always the "optimal" configuration for each source-package depending on which other packages are available. i use lfs only for servers where i don't need a gui. for desktops where the user has the choice to maintain the system himself i switched from debian to ubuntu. for a server the effort to solve the dependencies is ok for me, but for a full kde or gnome it's far beyond my capabilities. mybe i should hav written: i'll explain why *I* never will use any package-manager instead of i'll explain why i *NEVER* will use any package-manager thanks for your very informative mail tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: A Suggestion For A Simple Package Manager
Tushar Teredesai schrieb: > On Fri, Mar 20, 2009 at 3:01 AM, Tobias Gasser wrote: >> i'll explain why i NEVER will use any package-manager: >> > > Looks like you have based your comments based primarily on package > management used by "binary distros". The features that you describe > are available in source based package distros such as those used by > Gentoo, Lunar Linux. Gentoo uses USE flags to set what features to > compile for each package. Lunar Linux prompts the user the first time > an optional feature is available for a particular package. Both > distros (and I am guessing other source based distros) provide an > option to compile and install a minimal version of the package. sorry, but possibly we don't agree about the term "package management". i never thought about a source distro is using really a package manager. if i have to learn my sight is to narrow, be welcome to update my knowledge! for me, a package is something wich can be used to install something where as the package manager helps resolving all (or at least most) dependencies. the manager has options to remove or update a package too, maybe even an option to find unneeded packages. this definition (as i was convinced to be correct) implies the compile-time options for any package to be set by the package builder. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: A Suggestion For A Simple Package Manager
> This is a reasonable working definition from the viewpoint of the > user, it does not describe a package manager from the perspective > of the developer. finally that's it. you found the bug in my brain! i can follow your arguments and will try extend my knowledge about package management by googling arround... thanks for your explanations! tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: making LFS
RaptorX schrieb: > I wouldnt suggest to a newcomer to not follow the book but in general when > it comes to directories it is true that is a matter of preference... > > now you cannot forget about it because when in the book says go to > /mnt/lfs/tools you should go to the folder you created to replace that... > > and the cool thing comes with a long command that points to that folder > somewhere and you are not able to figure out why it didnt work, well it can > be that you forgot to replace the folder stated in the book by yours... ever had a look at http://www.linuxfromscratch.org/lfs/view/development/chapter04/aboutlfs.html the "$LFS" variable is used exactly to circumvent any problems with custom pathes for the tools directory. just the last branch "tools" is fixed. i myself never usede "/mnt/lfs", but i can live with the "tools" in my prefered location. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Kenel Panic
> now, I have a SATA drive 250GB, and my usb of 4GB. Those are the two > HardDrives that my pc should recognice. > CONFIG_SATA_PMP=y > CONFIG_SATA_AHCI=y you probably don't have the right sata driver. the only one you're building is the generic AHCI driver which supports most of the newer sata chips. to see which driver your slackware uses please perform the following steps: "lspci | grep SATA" on one of my systems i get the following 00:1f.2 SATA controller: Intel... 03:00.0 SATA controller: JMicron... now get the details by using the leading PCI device codes "lspci -s 00:1f.2 -v" the last line here tells me "Kernel driver in use: ahci" thus the CONFIG_SATA_AHCI fits for this controller "lspci -s 03:00.0 -v" result is "Kernel driver in use: ahci", thus bingo again! on one of my other machines i get "in use: sata_mv" for the onboard marvel controller. thus "CONFIG_SATA_MV" is required too. if you don't see any SATA, try it with "ATA" or "IDE". maybe your controller is mapped in the bios to behave as a standard IDE controller! check your bios setings and be shure to have it setup correct. sata must be native, not raid or whatever other options are available. ahci usually is only available in native mode. be carefull: changing the bios might result in unavailable disks from within other operating-systems installed! tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: Kenel Panic
> check your bios setings and be shure to have it setup correct. sata must >> be native, not raid or whatever other options are available. ahci >> usually is only available in native mode. > > > This should be optional right? yes. > this is in case I want to use the AHCI drivers... am I correct? no. my experience is: most bios map to pata when the chipset is not in native mode. you seem to bee lucky, as you have a bios mapping but the kernel still uses SATA_VIA. but i guess your via chipset does not support AHCI. more information you can find here: http://linux-ata.org/driver-status.html#via tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
kernel_sendpage - kernel 2.6.30.5
as the sendpage-problem seems to be a serious vulnerability in the kernel, i propose to add a paragraph in the errata-section and in the kernel chapters (5.6.1 / 8.3). in my opinion even an update to lfs 6.5.1 (or 6.6 is subnumbering is not acceptable) would be ok. i had no issues here building a new lfs with kernel 2.6.30.5. most changes from 2.6.30.2 seem to be minor fixes. just my humble opinion, but i'd like to read what you think about. tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: kernel_sendpage - kernel 2.6.30.5
Bruce Dubbs schrieb: > > In the All Packages section is a note: > > "The Linux kernel is updated relatively often, many times due to discoveries > of > security vulnerabilities. The latest available 2.6.30.x kernel version should > be > used, unless the errata page says otherwise." oops. i read the complete new book. but this paragraph i just couldn't remember... i guess i didn't sroll down as i already have downloaded all the packages i need to build the current system. > So I think we're covered. yes. but maybe to copy this note to chapter 5.6 and 8.3 would be a help for people with limited brain capacity like me (grin). tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Re: 5.7. Glibc-2.10.1
try the following: CC="$LFS_TGT-gcc -B/tools/lib/" \ AR=$LFS_TGT-ar \ RANLIB=$LFS_TGT-ranlib \ ../glibc-2.11.1/configure --prefix=/tools \ --host=$LFS_TGT --build=$(../glibc-2.11.1/scripts/config.guess) \ --disable-profile --enable-add-ons \ --enable-kernel=2.6.18 --with-headers=/tools/include \ libc_cv_forced_unwind=yes libc_cv_c_cleanup=yes that's what i had to do on xubuntu 9.10. the difference to the book are the three leading lines to point to the newly build tools (cc, ar, ranlinb) tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
5.7. Glibc-2.10.1
Chris Staub schrieb: > No...this is NOT necessary. If you "need" to add anything to the Glibc > build instructions, you have missed something. There is absolutely no > reason why the recently-built toolchain in /tools won't be used, if you > followed the instructions. If any variables "need" to be added, either > your PATH (or something in your user environment) is incorrect, or > something is wrong with the GCC or Binutils builds. nope. please try it yourself. just boot from the xubuntu 9.10 life cd, install the requirements (apt-get install gawk bison texinfo build-essential + symlink sh to bash instead of dash) and you'll see the glibc build failing. if you look at surajs message dated 2010/3/4 07.15 you can see the config output. the build and host system is recognised as "i686-pc-linux-gnu" (i now just extracted the "scripts/config.guess" and run it: same result here), the new built "i686-lfs-linux-gnu" toolchain probably does not fit and the hosts tools will be used (just my guess). tobias -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
Unable to compile GCC-4.4.3, Pass 1
Philippe Delavalade schrieb: > Maybe a suggestion for those who read to quikly : > five or six lines more in each chapter, something like > cd /$LFS/sources > tar -xvf package-version.tar.[gz|bz2] > cd package-version > ... > cd .. > rm -vfr package-version > and eventually > rm -vfr package-build > > It should perhaps prevent people not to follow 5.3 ? good idea. i'd go even one step further on. add another variable pointing to the packages. cd $LFS/sources tar -xvf $PCK/package-version.EXT cd package-version ... and with gcc cd $LFS/sources tar -xvf $PCK/gccX cd gccX tar -xvf $PCK/gmpX mv gmpX gmp tar -xvf $PCK/mpfr ... this avoids any relative path, everyting is fully specified by the variables $LFS and $PCK (or whatever you choose). greetings tobs -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
