-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Format: 1.7 Date: Tue, 23 Apr 2002 17:48:35 -0500 Source: pgi Binary: pgi Architecture: source i386 Version: 0.9.6.3 Distribution: unstable Urgency: high Maintainer: Progeny Debian Packaging Team <debian-packages@progeny.com> Changed-By: Branden Robinson <branden@progeny.com> Description: pgi - Progeny graphical installer creation system Closes: 142705 143792 143887 Changes: pgi (0.9.6.3) unstable; urgency=high . * High-urgency upload due to DFSG issues. * Preliminary PowerPC architecture support, care of Jimmy Kaplowitz. I fixed some comments and added Progeny keywords but this is essentially his work. (branden) * archive/Makefile: - define a no-op clean rule for the sake of the pgi-build clean change (see below) (branden) * busybox.h: - Turn off a bunch of busybox features in the live filesystem that we weren't using (we're using GNU ls and GNU sed, etc.). (branden) - Include busybox ls and more in the initrd; they cost only 4kB and greatly reduce the pain and frustration of working with the installer when the live fs can't be found (or has been unmounted). (branden) * debian/postrm: - Added postrm script to scrub /etc/pgi directory on purge. (branden) (Closes: #143792) * debian/rules: - stop shipping out-of-date RELNOTES file (branden) * doc/guide.xml: - New chapter: "Determining Package Placement on the Installer Media" (branden) - Document building system's requirement of loopback filesystem support. (branden) - Document target system's requirement of RAM disk and Initial RAM disk support. (branden) - Added "Packet Socket" to name of "Berkeley packet address family", since the former is what Configure.help calls it, and at least one person expressed confusion about this. (branden) * examples/configlets/do-apt-cdrom: - Center dialogs. (branden) - Display error dialogs if our apt and dpkg tricks fail. (branden) - Exit with a status of 1 if our apt and dpkg tricks fail. (branden) - Call the media "discs", not "CDs". (DVD-ROM, anyone?) (branden) * examples/configlets/extra_packages: - Install xloadimage as part of the configlets flavor. (branden) - Also install xfonts-{75dpi,scalable}. (branden) * examples/configlets/fvwm2rc: - Configure fvwm to decorate transient window and (hopefully) use a better looking default font. (branden) - Force the DeskTopSize to 1x1. (branden) * examples/configlets/options: - Removed. What the configlets need is now in woody and there are no other differences between this and the stock options file. (branden) * examples/configlets/postinst.sh: - Ensure that /etc/configlets is a directory, and install fvwm2rc, root_window.png, and root_window.xbm to it. (branden) * examples/configlets/xinitrc: - Don't run gnome-tasksel if do-apt-cdrom fails. (branden) - Set the X cursor to the wristwatch icon when starting, and try to set the root window to an image, if one is available. (branden) * images/{i386,ia64,powerpc}: - Use the --backup=numbered argument to mv when putting the ISO images in place. (branden) - (i386 only) Expand @PGI_VERSION@ when processing the syslinux text files. (branden) * images/i386/syslinux.screen08.txt.in: - Report PGI version on syslinux screens. (branden) (Closes: #143887) * {options,examples/configlets/options}: - Define the INITRD_FS and INITRD_FS_SIZE variables here (in the "private" section). (branden) * pgi-build: - Document the -P | --pgi-http-proxy option in pgi-build's usage message. (branden) - Define whether or not an architecture is experimental in pgi-build. Ordinarily this would go in the options file, but I think the package maintainer should have control over whether an architecture is considered experimental or not. (branden) - Store command line in a variable for later writing to generated ISO images. (branden) - Clean using the Makefiles' clean rules instead of the rm -r approach. (Closes: #142705) * python/pgi/base.py: - The test described in revision 1.55 was backwards, and so wasn't happening. After discussion with Eric, disabling the writing of sources.list in Stage 1 entirely. Eric points out that it's a Stage 2 job; for instance, base-config will clobber what we write no matter what, and configlets will have apt.conf one day soon. (branden) - Write trailing newlines to the initrd's /etc/fstab file. (branden) - When running debootstrap, delete PYTHONHOME from the environment altogether instead of setting it to null. (branden) - Force the exclusion of lilo from debootstrap's package list. lilo is extremely evil and does this unconditionally in its postinst: /usr/sbin/liloconfig < /dev/tty (branden) * stage0/Makefile: - Major overhaul and cleanup. + Remove old, unneeded variables, ifdefs, and rules. + Reorder the rules to make more sense. + Don't overwrite modules.dep files in the building system's /lib/modules even though depmod's design is so brain-dead that it's really tempting to do so. + More of my $@.tmp stuff to handle cases where a target may exist before we're done manipulating it. (This way, it doesn't.) + Make the clean rule actually do something useful. + Comment every rule of interest. + Stop defining INITRD_FS and INITRD_SIZE here, and s/INITRD_SIZE/INITRD_FS_SIZE/. (branden) * stage0/mkinitrd: - Reworked how we get kernel modules onto the initrd a little bit: Move the copying of kernel modules to be the very last thing we do to the initrd. Use du -ks to see how much space is actually taken up on it. . Then, for non-i386 architectures, subtract actual consumed space from the capacity of the initrd, and pass that figure to reducemodules.py. . For SYSLINUX i386, our initrd is extremely limited in size, because we have to squeeze it into a 2.88 floppy image. . Assume that gzip -9 can get us down to 60% of our original size, on average. Experimentation on actual kernel modules reveals that gzip does a little better than this. With all the text files on the initrd, this compression ratio should be better still. With this assumption, a full 8192kB filesystem would compress down to 4915kB. We need to be at most 2880kB, so our maximum allowed size is 4800kB, uncompressed. . In the future, ISOLINUX i386 can use the same method as ia64 and powerpc. (branden) - If $EXPERIMENTAL is defined, write its value to /etc/experimental on the installer initrd. (branden) - More sophisticated argument processing. (branden) - Quote variable contents that we write to etc files on the initrd. (branden) - Write the pgi-build command line and the options file used to the installer's /etc directory. (branden) - Look for the modules.dep file in $TDIR, where stage0/Makefile now places it. (branden) - Create tty[0-6] devices in the initrd. (branden) - Don't point /etc/mtab to /proc/mounts, since the system will try to write to it when me umount /proc before the pivot_root. (branden) - Recognize different name for kernel image files on PowerPC architecture, and add infrastructure for any architecture-specific differences in this respect. (branden) * stage0/stage0.sh: - Set $EXPERIMENTAL if /etc/experimental exists. (branden) - If $EXPERIMENTAL, howl a warning at the user and make them affirm their desire to continue; otherwise reboot immediately. (branden) - If $PGIDEBUG, launch a shell before performing hardware detection. (branden) - Remove the unconditional console chatter about the live fileystem. I've changed my mind about this; the logfile message is suffficient. (branden) - Write the pgi-build command-line to the log file, via /etc/pgi_cmdline. (branden) - If we succeed in killing processes or umounting /proc, break out of the re-try loop. (branden) - Export the pgilog function. (branden) - Unset the LOGFILE variable and pgilog function *after* their last possible use. (branden) * stage1/XF86Config.template: - Add UseFBDev option parameter to Device section. (branden) * stage1/Makefile: - Clean the live filesystem in the clean rule, not the distclean rule. (branden) * stage1/installer-druid: - General improvements to partitioning, originally in aid of Bugzilla#1091. Now all partition creation is done by make_partition, and no one knows or cares about extended partitions except it. (epg) - Uncomment the druid's logo-setting code, since an annoying bug in python-gnome has been fixed. (branden) - Use run_and_close() instead of run() in most situations, and document the one case where we really do want run(). (branden) * stage1/installer-text.sh: - Use the new unmount_targets() function. (branden) - Send standard error from debootstrap to the logfile as well. (branden) - Use a dialog "tailbox" to watch debootstrap's progress, and catch its exit status. (branden) - Remove now-bogus second argument to install_kernel(). (branden) * stage1/installer-x.sh: - Pre-answer the display/modes template to work around annoying debconf bug that is now fixed in "experimental". (branden) - Use the new unmount_targets() function. (branden) - Using the vga driver with "manual" monitor detection wasn't working. Fixed. (branden) - Set USE_FBDEV based on whether "cat /proc/fb" returns anything or not. (branden) - If debugging, launch shell after XF86Config is written, but before launching the X server. (branden) - Use debconf_poke on the xserver-xfree86/config/device/use_fbdev template with the value of USE_FBDEV. (branden) * stage1/{logoicon.xpm,root_window.png,root_window.xbm}: - Change Progeny logo (and nifty galaxy-based derivative) to Debian's Open Use logo, since Progeny is not comfortable licensing its logo under DFSG-compatible terms. (branden) * stage1/setup-finalize-text.py: - Remove reference to setup_cdroms(). (branden) * stage1/stage1.sh: - New shell function, unmount_targets(), that unmounts all filesystems mounted under /target. To be used by installer-*.sh scripts to get possibly-mounted target fs's unmounted before trying an installer script again. Harmless if no target filesystems are actually mounted. (branden) - Add logging to the debconf_poke() function. (branden) * stage1/xinitrc: - Set the cursor to the wristwatch icon as the very first thing we do. (branden) - Set the root window to solid dark slate blue unconditionally, before attempting to load an image file. (branden) * tools/detectmodules.sh: - Allow ethernet modules into the detection list. (branden) Files: 61b28412ab9ce011e5d719cfddef5623 650 admin extra pgi_0.9.6.3.dsc fc23ad6c7cff3284c943750d94da6666 480659 admin extra pgi_0.9.6.3.tar.gz dfd793bb93a28c8488560d4d1eb517e7 461240 admin extra pgi_0.9.6.3_i386.deb -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iEYEARECAAYFAjzF9SQACgkQ6kxmHytGonwbcACdFOjKNEVdTiN324bi2h/1eCu1 k5AAnRXF2eBwALvn4CXW3pk8IBsfA/N0 =qapV -----END PGP SIGNATURE----- Accepted: pgi_0.9.6.3.dsc to pool/main/p/pgi/pgi_0.9.6.3.dsc pgi_0.9.6.3.tar.gz to pool/main/p/pgi/pgi_0.9.6.3.tar.gz pgi_0.9.6.3_i386.deb to pool/main/p/pgi/pgi_0.9.6.3_i386.deb -- To UNSUBSCRIBE, email to debian-devel-changes-request@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
