source: MondoRescue/branches/2.05/mindi/CHANGES @ 124

Last change on this file since 124 was 124, checked in by bcornec, 15 years ago

add MINDI_CONF to the mindi LOGFILE (The warning that we use /etc/mindi/deplist.txt should be bigger, but that is a first step)
In that case mindi loops fdisk'ing -l /dev/ida/c*d*, including the
/dev/ida/c*d*p*, which is useless, but consumes an enormous amount of time.
The patch also includes some debugging enhancement
(Philippe De Muyter)

  • Property svn:keywords set to Id
File size: 44.5 KB
1MINDI                            CHANGES
31.05 (2005-11-19)
4- Bug fix for ldd output incorrectly handled, leading to "grep not found" error
5  (Andree Leidenfrost)
6- NFS now works in interactive mode, and nolock problems are solve
7  (Andree Leidenfrost)
8- IA-64 support is now working for rhel 3
9  (Bruno Cornec)
10- add MINDI_CONF to the mindi LOGFILE
11  (Philippe De Muyter)
12- Speed up fdisk'ing dev/ida raid devices
13  (Philippe De Muyter)
17- support exec-shield
18- added 'ide-generic' module to IDE modules in mindi to ensure that kernels
19  with fully modularised IDE sub-system boot
20- rewrote script 'wait-for-petris' to ensure that petris actually starts and
21  can be restarted reliably
22- fixed syntax error in mindi where a wrong delimiter is used in and sed call
23  when processing file '/etc/'
24- removed '#!/bin/bash' from file 'rootfs/etc/bashrc' (bashrc get sourced, not
25  executed)
26- removed executable flag from:
27  - 'rootfs/etc/'
28  - 'rootfs/root/.profile'
32- better support of SLES 8
36- better kernel-level logging
37- added ACL, xattr binaries to deplist.txt
38- fixed obscure bug which occasionally stopped mindi from correctly finding
39  and documenting all LVM2 LVM-on-RAID volumes
43- added ide_tape and other modules to mindi's config detection
44- unmount errant ramdisk ($mtpt) if fail to create boot floppy
45- better support of ISO dirs at restore-time (Conor Daly)
49- first 1.0x release
50- catch Ctrl-Alt-Del; trigger soft reset
51- better support of SuSE 9.1
52- added mdadm to deplist.txt
53- better detection of multiple Mindis (Martin Fürstenau)
54- don't complain if just a Mindi boot CD & not a platform for Mondo
55- updated busybox to 1.0.0pre10
56- removed uClibc
57- add memtest support
58- 2.6 kernel support
59- removed Embleer files (Andree Leidenfrost)
60- LVM v2 support for 2.6 (Takeru Komoriya)
61- added kernel-only floppy support, to accommodate really big kernels
62- updated+rebuilt busybox
63- added star support
64- mount /sys at boot-time
65- better 64-bit and 2.6 kernel support
66- better LVM, failsafe kernel support (Jim Richard)
67- use LILO, not raw kernel, on 1.4MB boot floppy
68- record names of unsaved modules for future reference
69- enlarged ramdisk by 8MB
72- changed some '==' to '=' --- now more RH6-friendly
73- allow absolute pathnames again in deplist
74- disable multifunc cd thing
75- better Gentoo support (Bill)
76- better OnStream support
77- better Slackware support (Laurenz)
78- added partimagehack-static to deplist.txt
79- recompiled Busybox - 10k smaller, better stack-handling
80- fixed boot screen typo
81- added support for 'auto' fs format
82- better devfs support for Mandrake users
83- better Debian+LVM support (Ralph Grewe)
84- updated analyze-my-lvm to handle floating-point gigabyte -L values
88- altered rootfs's /dev entry to stop cvs from becoming confused
90- added multi-function CD support to mindi and sbin/post-init
91- re-mount root as rw just in case
92- ask user to remove last data (floppy) disk if nec. (Tom Mortell)
93- added support for 5th column in mountlist.txt for labels
94- added symlinks.tgz
95- suppress erroneous error messages re: failsafe kernel
99- added cciss.o to SCSI_MODS
100- if format type is (e.g.) ext3,ext2 then use 1st entry
101- re-enabled fsck*
102- if cciss in use then enable it at boot-time
103- added /dev/ataraid/* to boot disk via ataraid.tgz (Luc S.)
104- better ISO support
105- tweaked mindi to use 10-15% fewer floppies
106- added RUN_AFTER_INITIAL_BOOT_PHASE var at start of Mindi, which
107  user may use to force Mindi to run command before bootstrapping
108  to aux data disks or CD - e.g. 'echo engage > /proc/scsi/something'
109- allow for Debian & other Stiefkinder that use 'none', not 'swap'
110  as the mountpoint of the swap partition
111- removed afio dependency
112- added RUN_AFTER_BOOT_PHASE_COMPLETE to let user specify a command
113  to be run by Mindi after it boots but before it runs mondorestore
114- re-worked and tarball not to use tgz's
115- cleared up the boot msg
116- updated busybox to 0.60.5; updated uClibc to 0.9.19
117- add #!/bin/sh to start of insert-all-my-modules
118- fixed obscure bug in install-additional-tools
119- change grep -m1 to grep | head -n1, for Debian users
120- moved lib.tar.bz2 and vmlinuz to mindi-kernel tarball/rpm
124- fixed LVM/RAID bugs (Brian Borgeson)
125- if bad lilo, give more verbose error before quitting
126- added mt and perl to deplist.txt
127- insmod ide-cd, cdrom, isofs, just in case
128- boot-time tmpfs ramdisk is now 40m (was 34m)
129- cleaned up logs
130- iso mode now calls Interactive
131- make SizeOfPartition() more Debian-friendly (Andree Leidenfrost)
132- clean up some calls to grep, esp. partition_mountpt=...
133- corrected some bashisms, to suit Debian ped- er, users
134- made first line refer to bash, not sh, to make sure
135  the Debian people know Mindi requires bash, not hs
136- changed grep -v "#" and grep -vx "#.*" to grep -vx " *#.*"
137- disabled code which would make Mindi use sfdisk instead of fdisk
138  if it looked as if Debian's fdisk would misbehave
142- updated ResolveSoftlinks() to work better with b0rkn Gentoo devfs /dev
143- patched analyze-my-lvm (Benjamin Mampaey)
144- detect built-in boot.b files in lilo
145- return w/err, don't abort, if Matt Nelson's RH8 system is farked
146- better at finding isolinux.bin
147- better at analyzing dependencies when running on broken distributions
148- try harder to boot from CD, even if tape fails
149- added ADDITIONAL_BOOT_PARAMS to be sent to kernel
150- better resolution of relative softlinks, leading to fewer
151  duplicates on data disks and therefore fewer data disks
152- detect Debian+devfs; use sfdisk instead of fdisk in that
153  eventuality, to work around yet another of Debian's warts
154- removed /lib/libuuid.* from rootfs.tgz
155- fixed vi
156- mindi now uses gawk --traditional (making gawk behave in a functionally
157  identical way to awk) - PASS; great, so now Debian needs to fix its awk :)
158- removed / from list of dirs accessed when trying to resolve deplist entry
159- spinner bugfix; handle odd LABELs properly (Tom Mortell)
164v0.90 (DEVEL)
165- new stable branch
166- EXTRA_SPACE=16384
167- fixed deplist.txt
172v0.72 (DEVEL)
173- misc code clean-ups
174- save boot device's boot sector
175- detect and beware Compaq diagnostic partitions
176- better handling of devfs V non-devfs kernels and boot devs
177- trimmed to busybox 0.60.3 binary on boot disk
178- mindi now resides in /usr/local/bin
179- removed lilo from dependencies
180- more Gentoo-friendly
185v0.71 (DEVEL)
186- if your kernel's builddate doesn't match any of the kernels in
187  your /boot directory then fudge the issue & find the closest
188  match (*grumble* Debian)
189- scan tape & CD at start, to force inclusion of modules on boot device
190- double EXTRA_SPACE if >7 disks
191- slimmer, more lithesome logfile
192- now accepts --findkernel
193- line 1982 - dd count=24000 should be count=$ramdisksize (Andras Korn)
194- better handling of non-Linux partitions on DevFS-enabled kernels
195- fixed Debian/ramdisksize/bloat problem (Johannes Franken)
196- fixed 2.4.20/fdisk eccentricity (Alistair Stevens)
197- updated kernel to 2.4.18-mdk6
198- generate mountlist in dev-abetical order ;)
199- fixed minor bug in .spec file
200- better at finding /boot/boot.b if your distro breaks the de facto standard
201- fixed minor bug in MakeMountlist
202- fixed
203- cleaned up deplist.txt
204- added rudimentary -v / -version flag
205- line 1180 or so --- duplicate mindi.iso entires --- fixed
206- added /bin/[ - a softlink to /bin/sh
207- ListAllPartitions() - sanity fix (KP)
208- fixed .spec bug
209- better feedback
210- added Markus's RAID patch
211- ListAllPartitions() - include /mnt/win* (Hugo)
216v0.70 (UNSTABLE)
217- better logging by dependency-calculating code
218- better handling of dependencies, specifically softlinks
219- if called by mondo then use mondo's temp dir as our temp dir too
220- changed gawk to $AWK in a few places, to allow for Debian
221- added host* config files to deplist
222- added tftp, ifconfig to busybox
223- added RPC support to uClibc
224- cleaned up creation of Mondo-Mindi configuration file
225- fixed bug in TryToFindKernelPath which stopped it from handling
226  multiple, same-version kernels gracefully
227- cleaned up deplist.txt
228- receive 'DIFFERENTIAL' variable from Mondo
229- better Debian compatibility, esp. w/detection of ver# (Hector Alvarez)
230- better devfs support; call fgrep in places instead of grep (Andrew Korn)
231- fixed analyze-my-lvm (Ralph Gruwe)
237- moved 50K of stuff from rootfs.tgz to aux-tools.tgz
238- added x11-tools.tgz option
239- updated /dev/console and /dev/tty0 (Paul Stevens)
240- cleaned up /mnt/groovy-stuff V /tmp/tmpfs code
241- catch sigint, sighup, etc.
242- fixed LILO-related message.txt mistake
243- .spec file clean-up; automation; config file (Carl Wilhem Soderstrom)
249- fixed RH7.3 readonly bug
250- improved nfs config file creation
251- search more locations for isolinux.bin
252- added isonuke option
253- faster data disk creation
254- better DevFS support (Hector Alvarez, DuckX)
255- nfs-related fix (Hans Lie)
256- abort if vfat filesystem present but mkfs.vfat missing
257- removed softlink to pico
258- allow Mondo to say no compression will be used
259- added [ to ramdisk
260- tar data disks with -b [block size] of 32k
261- don't autoboot to 'RESTORE' screen - it's scary!
262- let user choose lilo or syslinux as boot loader
263- added syslinux support
264- cleaned up message screens
270- improved logging
276- sped it up a bit
277- fixed ISO creation bug
283- fixed RAID mountlist bug (Daniel Grandjean)
284- better at German kbd support (Klaus Guntermann)
285- no longer abort if MONDO_PATH cannot be discerned
286- updated .spec file (Jesse Keating)
292- added less.bin to deplist.txt
293- copy mondo-restore.cfg configuration file to /var/cache/mondo-archive
294- increased EXTRA_SPACE to 16384
295- fixed bug in -H (RESTORE) function
301- added mkdir -p to fix mountlist.txt-copying bug (Bruno Vidal)
302- made sure mindi.iso (for tape users) copies kernel to [cd dir]/vmlinuz
308- minor cosmetic improvements
309- improved homedir-locating code
310- make sure the kernel is copied to [cd build dir]/vmlinuz
316- keep certain /dev directories compressed until boot-time, to allow
317  Mindi to support larger kernels
318- cleaned up logging a lot
319- better detection of the kernel currently in use
320- added /dev/nst* to ramdisk
321- amend isolinux.cfg to specify the correct ramdisk size
327- bootable CD's now use ISOLINUX instead of LILO (Andy Glass)
328- abort if isolinux not found
329- abort if multiple copies of Mindi found lying around
330- include mountlist.txt in ramdisk
336- change sbin/init's Die() function (Carl Wilhelm Soderstrom)
337- improve friendliness of kernel-related warning msg (Andy Glass)
338- added s/w switch re: making floppies (Carl Wilhelm Soderstrom)
344- excluded two unused vars (Carl Soderstrom)
350- reassemble large data/library files who have older 'mv' code which doesn't
351  like to overwrite symlinks - affects older RH+Slack systems (Andrew Bulhak)
352- call /usr/bin/logger, not 'logger', in LogIt()
353- run chmod 755 `which start-nfs` before running start-nfs
359- pathname problem fixed (Fred Feirtag)
360- patch to Debianize mindi a bit (H�tor Garc� �varez)
361- removed mondo-manual.html
362- moved /dev/{cciss,ida,rd} from rootfs.tgz to aux-tools.tgz
363- cleaned up the refs to analyze-my-lvm a bit (Hugo)
364- sped up disk-creation by 20-30%
365- make sure mondo-restore.cfg is on data disk as well as boot disk
366- added support for -C (cdstream) mode
372- quash error if /opt not found
373- use /tmp/groovy-stuff instead of /mnt/groovy-stuff for
374  handling data disks and their contents, for tape users
380- misc enhancements
381- uses self-resizing ramdisk (tmpfs) for /tmp
382- allows Mondo to exclude certain devices from mountlist
383- incorporated .spec patch (Ian Kent)
384- replaced &> /dev/null with > /dev/null 2> /dev/null
385- added vi to busybox again
391- fixed errant tar warning
392- replaced libc5 with uClibc
393- compiled busybox and open against uClibc
394- moved minor scripts+tools from rootfs.tgz to aux-tools.tgz
395- Mindi can now accommodate a kernel of up to 1290KB in size
396- no longer insists on cdrecord being present
402- more pointed error messages
403- fixed silly 'cp' error (Markus Marquardt)
404- use type C's embleer file for type B Windows partitions
405- increased EXTRA_SPACE from 4096 to 8192
406- cosmetic bugfixes
407- nfs users boot into nfs options by default
413- abort if mounted partition is specified as an imagedev
414- understand 'FAILSAFE' as kernel alternative
415- added analyze-my-lvm to tools available thru PATH
416- better discovery of locale and keymap file (Claude Mah�
417- compatible with 1.2x and the new 1.38+ stuff
418- better 'bug-me' script (Troff)
424- compatible with Mondo v1.5x devel branch
425- mindi's softlink now goes in /usr/sbin; was /usr/bin
426- include nfs.o, nfsd.o, lockd.o and sunrpc.o on data disks if available
427- patched .spec file to make it easier to change installdir (Ian Mortimer)
428- updated README
429- moved many global vars to start of script
430- better at finding US kbd maps
431- migrating to a single configuration file, /tmp/mondo-restore.cfg, for
432  Mondo users; file is not in use yet but is there for future expansion
438- MINDI_HOME and MONDO_HOME defined at start of each source file, making
439  it easier to relocate Mindi and Mondo if desired
440- tries to mount /dev/hd[a-h] at boot-time, looking for CD-ROM; was hd[a-d]
441- re-enabled the msg asking user to upgrade LILO if v21.6
442- Lin/Win users are much, _much_ less likely to run out of disk space now
443  when restoring from backups (Carlos Molina)
444- patch from Andrej Borsenkow to help w/Mandrake 8.1
445- disabled ramdisk free space checker (some distros don't like it)
451- don't abort if kernel is too big for 1.72MB floppy
452- better handling of eccentric distros and their keybd layouts
453- cosmetic fixes
454- sprinkled comments liberally inside deplist.txt
455- added gawk and awk to deplist
456- lilo has been added to .spec file
457- changed a couple of FatalError()'s to Die()'s
458- better TurboLinux 7 / LVM compatibility
464- better support of non-Red Hat, non-Debian keyboard resource files
465- cleaned up /sbin/init's output
466- replaced libc5 mount with user's own 'mount'
467- updated embleer.C.bz2
468- missing /dev/hdb2 found & replaced (thanks, Troff)
469- populated /dev/rd and /dev/ida on ramdisk (Michael Eisenberg)
470- improved SizeOfPartition() function
471- improved TryToFindKernelPath() function
472- don't offer NFS users Nuke Mode
473- finds and uses lilo.real if user is using Debian, a distro which
474  installs a shell script where the lilo executable should go
475- improved support for XFS users who opt to use failsafe kernel
476- uses user's lsmod, depmod, insmod, etc.
477- fixed freeze-up resulting from missing 'cat $mappath' command
478- mount tmpfs at /tmp/tmpfs for mondo-restore's benefit
479- try to create /dev/fd0u1722 if it is missing from /dev
480- added .spec file to tarball
486- fixed a bug in the handling of non-English keyboard maps
487- force failsafe kernel to provide reiserfs.o, xfs.o, jfs.o, ext2.o,
488  ext3.o and loop.o at boot-time, just in case
489- added /dev/sr[0-3] devices
490- copy LAST_FILELIST_NUMBER to /tmp, for compatibility with Mondo v1.3x
491- incorporates NFS-related settings, if supplied by Mondo v1.3x
492- replace busybox's mount with standard libc5 mount ater restoring
493  tools from floppies: NFS mounting requires full mount
494- always make mindi.iso
500- ramdisk's size has been tweaked
501- if script 'bootstrap' is found at boot-time then execute it
502- use tty8 to show mondo's output log
503- use tty7 to offer a game of Petris (if it exists on disk)
504- be sure to use i386-only keymaps
505- if LILO's "/tmp/dev.0" error occurs then abort & tell user to reboot
506- /sbin and /usr/sbin added temporarily to PATH when Mindi runs
507- if Mondo asks then force LILO to run in backward-compatible mode
508- if Mondo's tmpdir has post-nuke.tgz in it then extract it and
509  incorporate its files in the data disks
510- losetup /dev/loop0 -d before calling LILO (Norm Crowfoot)
511- if keyboard mapping dir (usu. /lib/kbd or /usr/lib/kbd) cannot be
512  found then don't abort; simply tell the user the default keyboard
513  mapping will be used
514- added /dev/hd[a,b,c,d][17-24] to the rootfs tarball
515- won't use LILO's optimizations when creating boot floppy; however,
516  optimizations are used on CD unless --broken-bios in call to Mondo
517- won't write data disks to start of tape; lets Mondo do that
518- tweaked how Mindi moves all.tar.gz around, to make it more
519  Mondo-friendly
520- tweaked Mindi's disk-splitter
521- user must unmount /dev/loop0 before calling Mindi: LILO insists...
522- added NFS-related entries to deplist.txt
523- if not called by Mondo then don't let user choose Interactive/Nuke/
524  whatever; just offer Expert
530    - removed libm.* from rootfs.tgz
531    - replaced sh with busybox's ash
532    - upgraded busybox to 0.60-2; pared it down, too
533        - include tape-related modules on boot disk, if backing up to tape
534    - creates a tape-friendly tarball containing tools+modules
535    - reassembly of tools+libraries is more reliable now
536    - streamlined deplist.txt
537    - give user their localized keyboard layout at boot-time
538    - fix some cosmetic bugs
544    - log Mindi's version# in the event of a fatal error
545    - make Mindi's boot-time output more esthetically pleasing
546    - if filesystem is type 'image', according to Mondo, then in filelist
547      its partition type will be listed as the hexadecimal string (e.g.
548          'b', '83', etc.) and the partition mountpoint will be 'image'
554    - support for newer features in Mondo v1.3x
558    - correctly lists RAID constituent partitions in mountlist
559    - breaks compatibility with Mondo v1.2x
560    - if user's fstab has weird line(s) then ignore; don't abort
561    - working on CDROM_MODS, trying to make it work on Stewart Charnell's
562      hardware setup
563    - cleaned up /sbin/init and /sbin/post-init a bit
567    - when insmod'g modules from ramdisk's root dir, use 'ls /*.o*'
568      (was 'ls /*.*') to find them
569    - if RAID controller(s) are detected then pause for 10s to let them
570      intialize by themselves (some 3ware RAID controllers require this)
571    - cleaned up /sbin/init and /sbin/post-ini; the former now controls
572      initialization of hardware, drivers, tools, etc. and the latter
573      runs mondo-restore if it exists
574    - better detection of and compatibility with SCSI CD-ROM drives
575    - chopsize is now 64KB; was 128KB
576    - cleaned up mindi's logging
580    - better logging of find-and-mount-cdrom's progress (to
581          /tmp/mondo-restore.log)
582    - somehow, lilo was re-enabled in deplist.txt; DISABLED (again *g*)
583    - added 'linear' to the list of modules Mindi always backs up
584    - if called by Mondo then pass the name & device of the user's boot
585          loader to the restore phase, by copying BOOTLOADER.* to the ramdisk
586    - if loading from floppies then try to mount CD-ROM a 2nd time after
587      extracting files from data floppies, just in case user is a CD user
588      who couldn't get their CD-ROM drive to work initially
589    - added fix to make sure /root/images/mindi is created (Charlie Root)
590    - aware of 'iso' boot mode, for future expansion
594    - added /dev/cdrom/cdrom[1-4] to the devices automatically excluded
595      from mountlist.txt
596    - added '-f-' to the end of most tar commands, to see if it fixes a
597      bug in some obscure Debian distros
598    - added network-related dependencies to deplist.txt
599    - if called by Mondo then don't write floppy images to disks; let
600      Mondo do that: less irritating for user
601    - if called by Mondo then do some clever tricks to add mondo-restore
602      and its dependencies (mondo-restore is now a dynamic executable)
603    - finally gotten klogd to squelch echoing of kernel msgs to console
607    - finally fixed that 'lib/modules/2.4.12-xfs not found' bug! :-D
608    - cosmetic changes and minor bugfixes
609    - excluded lilo.conf and lilo from deplist.txt; use user's copy instead
610    - if your fstab mentions a RAID partition but your raidtab doesn't then
611      Mindi will terminate in a sane manner with a sane error message
612    - fixed a bug in bug-me (heh, how ironic)
613    - include and run (sys)klogd to try to pipe kernel msgs to logfile
614    - call hdparm to speed up access to CD-ROM drive
618    - disabled fsck.* in deplist.txt
619    - include /etc/fstab in logfile
620    - removed lzop.* from aux-tools; should come from user's distro anyway
621    - tweaked data disk size and data disk formatting command
622    - make ISO bootable image (as well as floppies) when backing up to tape
623    - copy boot images (floppy + CD) from temp dir to /root/images/mindi
624      after making them, just in case the user wants them later
625    - added md5sum to deplist.txt
626    - added pico->vi softlink to ramdisk
627    - in the ramdisk's /etc/bashrc, added /usr/local/bin and
628          /usr/local/sbin to the PATH= command
629    - call insert-all-modules as background task; pause 3 seconds after
630      calling it, before proceeding <--- /sbin/init
631    - call to devfs has been moved to before call to post-init
632    - added 'md5sum -> busybox' softlink
633    - Mindi now uses busybox's printf, md5sum, mount, umount, vi
634      ...NB: I said, 'vi' :-) Now you have an editor again! :-)
635    - added lilo.real to deplist.txt, to allow for Debian's evil new distro
636    - added lzop to deplist.txt
637    - no longer including lzop in Mindi's tarball / RPM
638    - improved Mindi's error-reporting a little bit
639    - changed deplist.txt's /etc/* to just *
640    - restructured the calls in /sbin/post-init that reinitialize the LVM
641      and RAID modules
642    - if mkfs.ext2 is not in PATH then abort
643    - Mindi uses a 'filename cache' to speed itself up at run-time
644      (Jean-David Marrow)
645    - Red Hat 7.2 uses a weird /etc/; I've worked around it...
646    - tested with Linux-Mandrake 8.0 and Red Hat 7.2
650    - at boot-time, activate LV's if they exist
651    - restructured deplist.txt a bit (cosmetic change)
652    - re-enabled fsck.* in deplist.txt
653    - analyze-my-lvm fails gracefully if you're not using LVM but inserted
654      the module anyway
658    - made Mindi work with Red Hat 6.2 and 2.2.19 kernel
659    - if mkisofs is not found, say so; don't just say, 'an error occurred
660      while I was trying to make an ISO image'
661    - when formatting and writing floppies, if /dev/fd0's size is
662      indeterminable then I'll use /dev/fd0H1440 instead
663    - make sure xfs.o, jfs.o, reiserfs.o, ext2.o, ext3.o modules are
664      backed up (if they exist) to boot disks
665    - fixed a silly bug that stopped Mindi from handling its own modules
666      properly; only affects newbie users with broken kernels (i.e. the
667      very people that Mindi's failsafe kernel was meant to help!)
668    - added lzop (Marcus Oberhumer's file compressor) to aux-tools.tgz
669    - added Bruno Cornec's "--cd-recovery" patch; if user uses the switch
670      (i.e. '--cd-recovery yes') then the CD will boot to a screen that
671      pauses until the user types RESTORE; then, it will enter Nuke Mode
672    - added Stephan Zegherd's "superformat" patch
673    - failsafe kernel is now 2.4.12 with XFS, JFS and lots of SCSI modules;
674      the addition of XFS and extensive SCSI support will increase the
675      size of Mindi's RPM / tarball significantly but I think it's worth it
676    - at boot-time, find out whether there are any RAID partitions; and if
677      there are, whether they are 'started' yet; if not, run raidstart to
678      start them at boot-time, before user can try to mount anything
679    - Mindi's failsafe kernel & modules are handled more elegantly now;
680      if you have raid1.o, scsi.o, etc. loaded at backup-time then Mindi
681      finds its corresponding modules in its tarball of failsafe modules
682      and stores them so that at boot-time you can do SCSI, RAID, etc.
683    - removed gawk, printf from deplist.txt
687    - make sure lvm-mod, raid0, raid1 and raid5 modules are backed up
688    - added 'you need a ramdisk' to FAQ
689    - changed the slice size from 20KB to 64KB; speeds up Mindi by 25%
690      and doesn't have any negative side-effects (...that I can see)
691    - when searching for your current kernel version, do not examine
692      a file if it does not exist (it sounds obvious, I know...)
693    - when asking user to send me a copy of their log file, point out
694      that it may contain info which they do not want me to see but that
695      they are SOL unless I can actually see their log file
696    - fixed a silly bug that would crop up when a Linux-Mandrake 8.1 user
697      would use a devfs-enabled kernel and RAID partitions at the same time
698    - plays nicely with devfs-enabled kernels and distros
699    - data floppies are now 1.44MB each; boot floppy is still 1.72MB
700    - if binutils or libbinutils is not installed, abort
701    - fixed a bug which meant that some big files overwrote each other
702          at backup-time and caused library-related problems are boot-time
703    - enabled comments ('#') in deplist.txt
704    - disabled fsck.* in deplist.txt; user may re-enable the line if
705      s/he wants those tools available at boot-time
706    - tested successfully with the following distros
707      - Linux-Mandrake 8.1 and its 2.4.8, devfs-enabled stock kernel
708      - Red Hat 7.1 and its 2.4.2 stock kernel
709    - better at guessing the size of swap partition
710    - drop minix tools from deplist; user can restore them if s/he wants
711          to use the minix filesystem
712    - don't backup _all_ cdrom modules (if backing up 'cdrom.o'); be
713      selective! ...minor bug in ListKerneModulePaths has been fixed
714    - if the user has two glibc libraries - one in /lib and one in
715      /lib/i686 - then Mindi will re-jig the softlinks so that only the
716      version in /lib is backed up, to save space on the data disks
717    - major rewrite of the code that copies the tools and libraries from
718      the data disks; prettier, faster and more stable (and did I mention
719      how kind it is to your hard-working hands?)
720    - better handling of crazy libraries and softlinks thereto
721    - new boot-time script, 'bug-me', copies user's logs to a floppy and
722          tells him/her to e-mail them to me if mondo-restore aborts
723    - ignore kernels named vmlinux (note the 'X'); traditionally, kernels
724      are named vmlinuz (note the 'Z') or something similar
725    - no longer backup all modules to CD/tape; just backup the ones loaded
726          at backup-time; this saves boot-time ramdisk space
727    - removed all references to LOOPDEVICE
728    - LVM is supported (experimental; watch this space...)
729    - included a tarball of 'sane' /dev/loop* nodes which will be extracted
730          and added to the boot disk; this helps maintain Mindi's backward-
731          compatibility with non-devfs kernels and distros
732    - when copying modules to boot or data disks, uncompress them first
733          if they are gzipped (some distros come with compressed modules)
734    - no longer copies all loaded modules to the 2.88 MB El Torito bootable
735      floppy image; instead, it copies FLOPPY_MODS to the 1.72 MB image,
736          CDROM_MODS to the 2.88 MB image, and _all_ modules that were loaded
737          at backup-time to the data disks
738        - supports tape streamers, in conjunction with Mondo
739    - uses busybox's sync again; let's see if that breaks anything...
740    - if Mondo is backing up to tape streamer then write the tape's size to
741          /tmp/TAPESIZE and the drive's device to /tmp/TAPEDEV-LIVES-HERE on
742          the ramdisk
743    - floppy disk modules (ide-floppy.o and floppy.o) now are copied
744          to the ramdisk, whether the user's kernel or the failsafe kernel
745          is being used
749    - excluded *.old from the 'boot.b' list when searching for boot.b
750    - updated Mindi's own failsafe kernel+modules to v2.4.10
751    - added umount to deplist again
752    - improved Mindi's support of modular kernels (esp. modular CD-ROM
753          drive and floppy disk drives)
754    - don't accuse Linux 2.4.10 users of having a 'buggy kernel'
755    - restored the libc5 versions of sync and umount (dropped from
756          deplist but also from busybox)
757    - fixed a lot of silly bugs related to incorporating Busybox in Mindi
758        - experimental LVM support
759    - added 'analyze-my-lvm' script to the tarball; Mindi uses it now
760    - able to recognize & record LVM mounts, adding them to the mountlist
761    - added the LVM executables (listed in the lvm-0.xx-i386.rpm) to
762      deplist.txt; they won't be added unless you have them and you won't
763      have them unless you're using LVM, so this is not 'bloat' IMO :)
764    - added date and sync to deplist again
765    - when working on LVM stuff, use 'echo y | pvcreate -ff' (was '..-f')
766    - amended format-...-kludge to allow for mountlists with >1
767          spaces between cols
768      checked the fsck.reiserfs is in deplist.txt (yes it is)
769    - incorporated busybox 0.60-1; now, Mindi is faster at run-time and
770          smaller at boot-time; you can fit a typical system onto 3 floppies;
771          also, a kernel up to 1MB in size (approx.) can now be accommodated
772    - supports kernels that have MODULAR support for CD-ROM's or floppies
773    - Mindi copies only the loaded modules to the data disks but it copies
774          _all_ modules to the CD, just in case
775    - thanks to busybox your floppy-based Mindi boot disk set should now be
776          able to accommodate a kernel of up to 1MB in size
777    - boot into Expert Mode unless boot disk was generated by Mondo
778    - exclude all softlinks from list of 'eligible' kernels
779        - exclude kernels with 'vmlinux' in their names
780    - removed A LOT of files from deplist.txt's dependency list; they are
781          listed in deplist.removed, in case you want to put them back :-)
782    - excludes libX11, libXext, libXi, libgtk, libgdk ALWAYS, whether you
783          add them to deplist.txt or not; [cancel this by editing line 485]
784    - rewrote AddFileToDir() to make it more efficient (but slower...)
785    - strip chaff from binaries; strip comments from bash scripts; strip
786          comments from /etc/termcap
790    - if you use a softlink to your CD dev, e.g. /dev/cdrecorder->scd0,
791          then Mindi should handle it properly now (i.e. not bomb out)
792    - if you have a large kernel & lots of modules loaded, Mindi will only
793      copy as many modules to the ramdisk as will _fit_ on a ramdisk that
794          will have to go alongside your kernel on a 2.88MB floppy disk
795    - if your kernel does not support loopfs, Mindi will say so & abort
796    - turned some warnings into fatal errors, to help the user to figure
797      out exactly why Mindi won't work on their kernel
798    - changed mke2fs -N24 to -N32, to see if it helps braindead distros
799    - Mondo now tells Mindi whether or not to use its stock kernel
800    - gone back to the old 'insert-my-modules' for now; it's more reliable
801    - working on the 'insert-all-my-modules' script, to make it compatible
802      with a wider range of insane distros and kernels
803    - dropped the '-c' option from LILO call
804        - removed cdrecord, mkisofs, vi, route and diff from deplist.txt
805    - slowed down the 'split data between floppies' routine; made it
806          try harder to squeeze more data onto each floppy
807    - excluding nvidia modules from the "insert-all-my-modules" script
808          because it sometimes causes freeze-ups
809    - added code to copy your fstab, lilo.conf, etc. to a tarball which, in
810          the event of an error, Mindi will ask the user to e-mail me
814    - plays nicely with devfs, so long as your old /dev/hdX, /dev/sdX
815      softlinks are still there :-)
816    - if your swap partition isn't mounted, Mindi will decide that it is
817      125MB; it used to abort but now it just warns & makes a good guess
818    - testing the failsafe kernel & associated scripts
819    - only use lilo's "-c" option if NOT a 2.88MB floppy disk image
820    - copes better with .o, .gz, etc. module suffixes
821    - better 'loaded modules / not loaded modules' feedback at boot-time
822    - semi-forces isofs and ide-cd (not isofs.o and ide-cd.o); the former
823          will be more compatible with stock kernels - e.g. Red Hat, Mandrake;
824          also added vfat, fat & a few other modules to the list of 'foced'.
825    - removed the '-c' switch from the call to LILO (some BIOSes don't
826          like it very much)
827    - removed chaff from ramdisk's /etc/login.defs, saving approx 7K
828    - format the boot disks with -N 24 (was -N 64)
829    - do not insist on CD #1 (at boot-time) if there is no 'CD #' marker
830    - amend format-and-kludge-vfat; now, it will not abort if it cannot
831      find its entry in mountlist; it doesn't need to find the entry
832          anyway, except to check that the kludge works (which it does)
833    - if user chooses to use Mindi's failsafe 2.4.9 kernel, Mindi will load
834          Mindi's versions of user's modules at boot-time
835    - make label-devices-as-necessary log its efforts
836    - added e2label and tune2fs to deplist.txt
837        - if you boot in Normal Mode, I'll call mondo-restore --interactive
838          instead of just mondo-restore
839        - replaced K7-compatible 2.4.7 kernel with a 586-comp'ble 2.4.9 kernel
840    - improved the post-run clean-up
841    - force the copying/loading of isofs.o and ide-cd.o, which enhances
842          Mindi's support for kernels that have modular CD-ROM support
843    - Interactive Mode offers to edit /tmp/mountlist.txt before running
844          mondo-restore (good idea from F. Boon)
845    - /dev/hd[e-h]* have been added to the ramdisk
846    - disabled Cuckoo and Bootstrap Modes (they weren't doing anything
847          anyway & were probably confusing some users)
848    - removed 'edit mountlist?' question (because mondo-restore takes
849          care of that now, via a user-friendly mountlist editor)
850    - included a copy of mondo-manual.html
851    - only kludge /dev/hda1 or /dev/sda1 (if vfat, of course);
852          otherwise, use regular mkfs.vfat
853        - use mkfs.dos if mkfs.vfat does not exist
854    - moved some scripts from rootfs.tgz to Mondo's restore-scripts.tgz
857v0.38                                                              08/18/2001
858        - clarified the 'more than one kernel found' comment
859    - updated the README to emphasize that the scripts are deprecated
860    - when displaying mountlist on screen, make it easier to read
861    - changed most 'echo's to 'LogIt()'s; log most screen output to
862          /var/log/mindi.log
863    - user can now use a stock 2.4.7 kernel (which comes with Mindi)
864          if the user's distro's kernel isn't right for a boot disk
867v0.37                                                              08/12/2001
868        - changed the boot screen a little bit
869    - changed the "cannot boot from CD-ROM" message (more informative now)
870    - if an 'insmod'ed module cannot be found, warn but don't abort
871    - that whole loop-mount/extract tgz to ramdisk thing --- if it fails,
872          abort, don't try to continue; also, ask if loopfs is supported by
873      kernel (if not, that might explain the error)
874    - in addition to /lib/modules/`uname -r`/modules, also search
875          /root/oss/modules for kernel modules (Phillip Deackes)
876    - run syslog and klogd at restore-time, to pipe sys & knl messages to
877          /tmp/mondo-restore.log
878    - default restore-time boot mode is now 'interactive' (unless the
879          disk is a 1.72MB, in which case default mode is still 'expert')
880    - if CDROM can't find itself (long story, re: El Torito) then the
881      script will say, "Your kernel is b0rken." This will reduce the
882      number of erroneous error reports that I receive from users
883    - working on implementing Cuckoo Mode again
886v0.36                                                              08/03/2001
887    - testing it with LM8, RH71 and Slack8
888    - explicitily exluding /mnt/floppy and /floppy from mountlist.txt
889    - warn if afio not found (Mondo needs afio)
890    - optional --custom <temp dir> <out dir> switch added, to let the user
891      specify where the temp dirs and created images will be stored
892    - changed /sbin/init to install modules _before_ trying to mount
893          CD-ROM (David Granz)
894    - create a gawk->awk shortcut at restore-time if necessary
895    - specifically exclude /dev/fd0h1440 and /dev/fd0H1440 from mountlist
896    - removed silly "was CD inserted?" message
900    - if LILO fails, it writes its errors to the screen
901    - use user's boot.b and LILO instead of Mindi's
902    - tarball/RPM no longer includes boot.b or lilo
903    - mke2fs' output will be echoed to screen, if an error occurs
904    - doubled the -N inode allocations and removed some of the -i inode
905      fiddles, just in case they are making Red Hat 7.1 choke
909    - includes patch to rid Mindi of its dependence on /dev/loop devices
910      (Bill Davidsen)
911    - includes patch from 0ystein Skalsem, to fix a disk-formatting problem
912    - warns if kernel is 2.4.0-2.4.5; encourages user to upgrade
916    - excludes all devices containing ':/' (i.e. NFS mountlists)
917    - the FORMAT/SYS replacement, "format-and-kludge-vfat" now writes
918      its logfile to /tmp/format-and-kludge-vfat.log
919    - /etc/fstab now mounts floppies for read/write access by default
920    - format-and-kludge-vfat should only write to log once now
921    - added vi, top to deplist.txt
922    - updated the rootfs.tgz's README
923    - mountlist is written to screen as well as mountlist.txt
924    - tested with Linux 2.4.7; works fine
925    - changed 'mkfs' lines, to format the loop-mounted files instead of the
926      device nodes which makes Mindi work better with kernels 2.4.6+
927    - Mindi can now handle kernels of up to 930K in size
928    - sped up the process of spreading data files across data disks
929      by approx 20%
933    - software RAID is now supported
934    - added the 'raidtools' binaries to deplist.txt
935    - add /etc/fstab and /etc/raidtab to deplist.txt
936    - added /dev/md* to the ramdisk's /dev directory
937    - modified MakeMountlist() to calculate correctly the sizes of
938      RAID partitions, if they are present in /etc/fstab
939    - moved the 'LOOPDEVICE=...' code into the regular main() section
940    - install-additional-tools makes a softlink '/etc/raidtab' that points
941      to /mnt/groovy-stuff/etc/raidtab; also, '/etc/fstab.orig' that points
942      to /mnt/groovy-stuff/etc/fstab
943    - added line to 'install-additional-tools' that will delete the
944      dummy modprobe before installing data from datadisk(s)
945    - mkdir -p $outdir at the beginning, just in case...
950        - included DoJ's patch to make Mindi play more nicely with devfs
951    - slightly better and README
955        - added klogd and syslogd to deplist.txt
956    - can now cope with Red Hat's "LABEL=" entries in /etc/fstab
957    - should ignore NFS-mounted partitions now
961    - stopped the 'modprobe' errors that used to occur at the beginning;
962      they didn't mean anything but they did worry some novice users
963        - removed losetup from rootfs.tgz (wasn't doing anything useful)
964    - removed the 'txt-or-gui' stuff from Mindi (never used it anyway)
965    - added clear to deplist.txt
966    - moved rootfs.tgz's "mondo/" scripts to Mondo's "restore-scripts.tgz"
967      tarball; tarball will be untarred and contents added to the
968          data disks at run-time, if Mindi finds the tarball
969    - if user lacks fd0u1722 but has fd0h1722 then use the latter
970    - boot-up msg is more informative
971    - fixed obscure bug in find-and-mount-cdrom
972    - better at coping with hda1/2/3/4 (normal people use hda1/5/6/7 but
973      if you really want to exhaust your primaries, well hey, your life)
974    - the 'outstring sucks!' message is now more informative
978        - if AddFileToDir() is called with a file that does not exist, Mindi
979          warns but does not abort
980    - should work with gzipped modules (which end in '.o.gz') and
981          NVidia modules (which don't end in anything)
982    - should work with Red Hat's now
983        - misc. cosmetic changes to Mondo boot-time and restore-phase scripts,
984          to make the visual output cleaner & more legible
985    - insist-on-cd retracts the CD tray after asking for next CD, in case
986      the user did not close the drive before pressing <enter>
987    - if unable to create 1.72MB boot disk then warn but do not abort
988    - put creation date on the boot-up screen
989    - create data disks with tarballs up to 1700KB in size; tries harder
990          to fill each data disk to the brim
991    - changed the call to make ext3 partitions; now, it uses a call to
992      mkfs -t ext3 -j -Jsize=10 -q <device> (as requested by ext3 manual)
993    - tested w/ext3; works fine
994    - install-additional-tools doesn't call sync until after all softlinks
995      have been finalized; that's to make sure we call the _right_ sync
996    - ramdisk is now (size of all files in data disks PLUS 8192) KB
997    - RPM and SRPM published
998    - mount CDROM at start of untar-me, just in case it wasn't mounted
999    - when restoring selectively, log the archives that have matches
1000          & the number of matches (inc. if zero)
1004    - has been tested with XFS, JFS, ReiserFS, ext2 and vfat
1005    - compatible (but not tested) with ext3
1006    - better at finding your kernel, even if you have several kernels
1007      of the same version (now checks the build date, too)
1008    - better file-spreading algorithm; a little faster; much more reliable
1009        - includes a copy of LILO, to make the boot disks properly
1010        - tells you how big the tarballs are, when it assembles them
1011    - compare-me uses 'insist-on-cd' instead of internal subroutine
1012    - it looks as if Mindi can actually cope with 920K kernels now :)
1013    - fixed some silly multi-CD bugs that would stop big files from being
1014      restored if they were spread across CD's
1018    - uses its own copy of lilo
1019        - removed all "~" (backup copies) from rootfs.tgz
1020        - updated the rootfs's README - stablilo-me, not hack-me
1021    - improved the formatting of the lilo.conf as it is created
1022        - replace all libc5 tools with libc6 (user's) tools at restore-time
1023    - added diff to deplist.txt
1024    - copy io.sys and msdos.sys (if found) to data disk(s) at
1025          run-time; softlinks them to /*.sys at boot-time
1026    - fixed some vfat formatting-related issues w/Embleer (always tricky);
1027          the kludge's error-handling is better and its exit is more graceful
1028    - say 'type nuke to restore automatically' in msg if Mondo called me
1029    - backup all modules, not just the ones present/loaded at run-time;
1030      the loaded modules go on boot disk but ALL go on data disk
1031    - reduced the max data per disk from 1700 to 1680, to suit ext2's needs
1032    - DidMondoCallMe() replaces 'ps ax | grep | blah blah'
1033    - includes user's /etc/termcap and /usr/bin/pico on disk(s)
1034    - cosmetic changes to restore-time scripts (spaces, newlines, etc.)
1035    - pipe kernel's msgs to /var/log/messages
1036    - make sure /etc/lilo.conf is 600 (RW by root only)
1040    - can accommodate a kernel >900KB in size
1041        - generates lilo.conf on-the-fly; no default/template file anymore
1042    - copies io.sys and msdos.sys to the CD (if you have them)
1043    - add the embleer tarballs to mindi-0.xx.tgz, so that users can
1044          format and make bootable msdos/Win9x partitions
1045    - added printf and hdparm to deplist.txt
1046    - incorporated Bruno Cornec's Mondo 0.991 scripts in Mindi
1047        - separated README into README, INSTALL and TODO
1048    - fixed silly {ramdisk}/sbin/InsistOnCD bug
1049    - took out mke2fs and libext from rootfs.tgz
1050    - rootfs.tgz's "/etc/bashrc" now includes /usr/share/mindi
1051    - compare-mode uses IDE optimization to speed things up
1052    - don't quibble over resizing partitions if <2MB difference
1053    - fixed silly InsistOnCD bug
1057        - don't copy *.mdsg to boot floppy image
1058    - fixed erroneous 'cannot make ISO image' error message
1059    - can accommodate a kernel as large as 890K
1060        - now uses lilo, not syslinux (lilo works better w/2.88MB boot images)
1061    - creates & formats disk images on the fly (no more .img.gz templates)
1062    - boot disks are formatted ext2, like the data disks
1063    - creates 1.72MB and El Torito 2.88MB boot floppies/images
1064        - if called by Mondo, doesn't list images or offer to copy to FDD/CDRw
1065        - includes the distro's name and kernel version in LILO's message
1069        - don't add a ramdisk; use rootfs & that's it
1070        - dependency-calculator runs approx. twice as fast as in v0.20
1071        - tested with Linux Mandrake 8.0; works well, if you recompile
1072          kernel to include floppy disk support (which LM8's kernel does not)
1073        - supports bigger kernels now, up to 900K in size
1074        - deplist.txt may now include directories, not just files
1075        - ignores kernel filenames that are hyperlinks
1076        - uses awk instead of gawk; drop the awk/gawk-finding thing
1077        - uses gzip, not bzip2; dropped bzip2, mv, ls to make rootfs smaller
1078        - copies user's /dev/fd0* devices to boot disk
1079        - calls Mindi's copy of syslinux, not user's copy
1080        - use modules' paths, e.g. instead of just 'insmod fubar', use
1081          'insmod /lib/modules/2.4.5-ac117/vfat/fubar.o'
1082        - in boot.msg, XXXXX is part 1 of the 2nd line, YYYYY is part 2 of
1083          the 2nd line, and ZZZZZ is Mindi's version# (in the 1st line)
1087        - df, du, find, lsmod, rmmod have been removed from rootfs.tgz
1088        - at boot-time, inserts kernel modules w/user's original parameters
1089        - deletes the slices after recombining them & restoring the files
1090        - includes the MINDI_VER version number in boot.msg
1091        - if boot disk is too small, lists the files & which one is too big
1092        - syslinux now waits 10 seconds before booting floppy
1093        - 'insert-all-my-modules' is copied to rootfs, not to a data disk
Note: See TracBrowser for help on using the repository browser.