Luca wrote: > Luca wrote: > >> Luca wrote: >> >> >>> Alexander E. Patrakov wrote: >>> >>> >>> >>>> Luca wrote: >>>> >>>> >>>> >>>>> Hi! >>>>> >>>>> Some days ago I found and started using T2 Open System Development >>>>> Environment scripts (http://www.t2-project.org) to build a system and >>>>> Live/Installation CDs. >>>>> >>>>> >>>>> >>>>> >>>> According to the information on that site, the developers did almost >>>> the impossible: crosscompiled Xorg. So the approach looks promising. >>>> >>>> >>>> >>>>> Now after testing again the system, I'll begin to modify the scripts and >>>>> then talk again with Rene; after the lfs projects are built using these >>>>> scripts I'll put them into tests to see and check the results, failures >>>>> etc... and then if he said that there are no problems, if someone is >>>>> interested I could post the original and modified tarball. >>>>> >>>>> >>>>> >>>>> >>>> Do I understand correctly that, once the scripts are created, they >>>> have to be manually maintained and updated, so that changes that >>>> happen in the books are propagated to the scripts? >>>> >>>> >>>> >>>>> The idea was to modify scripts to build the stable or svn/development >>>>> versions of the books; something like ALFS but with the opportunity to >>>>> create even CDs/DVD Live and Installation images. >>>>> >>>>> >>>>> >>>>> >>>> This also has the value of build-testing BLFS against CLFS and >>>> identifying problem spots. So go ahead, we'll see the results. >>>> >>>> If you want to get notes on i18n, RW support, and release goals of the >>>> past CD, mail me privately. >>>> >>>> >>>> >>>> >>> Hi Alexander, >>> >>> yes, you understood correctly, the scripts need to be maintained and >>> updated by hand, but I'll try to fix this (to update the original >>> T2-<version> it's possible to add a cron job that download the sources >>> from svn). Actually I began to modify the scripts to take care of LFS >>> (I'll try to add additional configuration options too), but I don't know >>> exactly when I'll put on fire the modified scripts (it takes a lot of >>> work and until now I modified everything about system and configuration >>> - but only for x86 for now to test - and autoconf, automake, bash, >>> berkeley-db, binutils, bison, bzip2 and coreutils files from >>> LFS-SVN-20061008 book. >>> Since the first stages are cross-compiled I'll take a look to see how to >>> patch packages only for that stages, and not in the final/native system >>> (the way CLFS does with the posix compliance patches). >>> I think that the result should be a mix of CLFS and LFS, not exactly the >>> pure standalone books... >>> >>> Thanks, if I need something related I'll email directly to you, however >>> I'll post from time to time on the list the work done. >>> >>> Not sure, but from what I read, if I understood correctly, it should be >>> possible to create a DVD image too with these scripts. >>> >>> Luca >>> >>> >>> >> I put the scripts on fire although for now they build only bare lfs >> system x86 version, just began stage 5 of compilation, the first 4 >> stages built without problems but I'll wait all the remaing stages to >> begin to trace out all logs (until now 40MB of log files). >> >> Luca >> >> >> > Just finished to run the scripts and created the 2 CDs set Installation > images after 2 days of compiling. > Finished the 7 stages compilation and checked through 150M of logs. I > had only 7 packages that failed, in stage 5 (blender, embrace - e17, > gnupg2, koffice, qemu, transcode and vnc), which I'll inspect. > > Now, after burning the isos I'll install the system (full lfs/clfs plus > blfs and various extra software) and start testing. > > Luca > Hi!
The blender problem was solved with a patch pointing to Python headers, gnupg2 was solved by adding gnupth (not a mandatory dependecy but required) to the list of packages to be installed and transcode solved adding glib12/gtk+12 to the list of packages to be installed, the others I'm still investigating. After that, I burnt the two isos and began installation (I've actually not added support to reiser4, but it is only a question of two patches and a package to be added); I created my partitions on a blank hard disk, created filesystems, installed everything and had no problems (only vim-7.0 perl bindings which broke during make, the same way they do while building lfs/blfs). Solved the problem with applying patches only in toolchain/cross stages; actually I'm taking care about adding gnat in the toolchain stage so to have at the first gcc compilation an ada compiler ready. Alexander, I have a question: (the native stages began in stage2 through stage9), glibc (or the option given by HLFS uclib, or eventually another option dietlibc, needed because some packages needed to create the iso, example embutils, get linked against) takes care of the locale setting (which we actually export in chapter 7 of lfs book and chapter 3 of blfs book); how much is it safe to create the /etc/profile and /etc/profile.d/i18n.sh with language settings after installing glibc in the first stage of native system? Is it better to wait until all stages get compiled? Just a note: every package compiled with these optimizations: "-02 -pipe -s -fomit-frame-pointer --target=i686-* --host=i686-* -march=athlon-4", so final gcc command (in native stages) looked like: "gcc -pipe -02 -march=athlon-4 --version -s -fomit-frame-pointer" Actually since my cpu is 64bit one, even if for some systems I use it as 32, I could try a x86-64 multilib or pure 64 build; I could try, but didn't know much about, cross-compiling between archs and then testing with some emulator (bochs, qemu, or whatever). Bye and thanks, Luca -- http://linuxfromscratch.org/mailman/listinfo/livecd FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
