Bug#514016: options for fixing

[Thomas Viehmann]

On 2009-02-06 09:57:36.00 Raphael Hertzog  wrote:
> On Thu, 05 Feb 2009, Thomas Viehmann wrote:
> > > What would dpkg do with the symlinks in the the unpack phase?
> > 
> > I have not tested it, but I would venture the same thing (calling tar,
> > ergo having the same problem) from when I last looked at it.
> 
> dpkg does not call tar, it has it own tar implementation (which explains
> why we don't support yet POSIX tar archives inside .deb). I'm not sure how
> it behaves in the situation that matters for this bug.
> 
> Try it with dpkg --root=/target/dir/ …

Thanks for stepping in to correct me here. dpkg does have its own tar, only 
dpkg-deb calls tar.

> Try it with dpkg --root=/target/dir/ …

>From looking at the code, it does the right thing.

Kind regards

T.



-- 
To UNSUBSCRIBE, email to debian-boot-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#514016: options for fixing

[Thomas Viehmann]

Hi,

thanks for your replies. Again, this is mostly because I needed some
idle occupation and am not into sudoku. Apologies for presenting it a
way that looked like a proposal.

Bdale Garbee wrote:
> How about just fixing the offending package to use a relative symlink,
> like /lib64 -> lib?  This is well within policy since they're adjacent
> objects in the same directory, and is less likely to break things or be
> a pain to maintain over time than any of the options above.  I can't
> think of any negative consequence to this change, offhand?

Well, having 1000 maintainers do the right thing is difficult, as proven
by the release process.

> Adding a Debian-specific option to tar would certainly not be my first
> choice.

Other options do exist, my list probably is not exhaustive, but I would
think that they cost a lot of code with a lot of opportunity to add bugs
where things work now and it might be interesting to have an upper bound
for how involved a minimal fix is.

Thinking on a larger scale, what we actually want to do here is unpack a
tarball into a chroot. I venture that this is a reasonable use case of
tar and that it might just be possible to try to have upstream agree to
have such an option in principle so that there is a plan to get rid of
the debian-specificity of the option.

Bastian Blank wrote:
>> - switch off tar,

> I intend to replace tar with a custom unpack routine in cdebootstrap.

TBH, I'm not that impressed by the implementation of ar (or a subset) in
dpkg and am not sure that reimplementing tar extraction it is a good
path to take when I would think that an option for tar to do the right
thing (whether internally munging symlinks when resolving paths or using
chroot) might have a general use beyond bootstrapping Debian systems.

>> - try to do something with tar
>>   (new upstream - vs. release: ugh - has a "apply to symlink target"
>>--transform flag, but that also might need post-processing of
>>symlinks at end of bootstrapping),
>
> What would dpkg do with the symlinks in the the unpack phase?

I have not tested it, but I would venture the same thing (calling tar,
ergo having the same problem) from when I last looked at it.

>> - try to do some ld-preload(?) trickery for changing the way symlinks
>>   are read,

> fakechroot have to do this someway, but I was unable to grok that code
> yet.

Well, you'd have to intercept all the relevant calls (for
dereferencing/reading symlinks). Seems doable, but IMO tar would be the
better place. But if the masses want funny business, so be it.

...
>> - add chroot option to tar (see patch).

> This is a root only option.

Yes. It was my impression that (c)debootstrap are supposed to be called
by root. At least I always used to call them as root when I was a developer.

Kind regards

T.
-- 
Thomas Viehmann, http://thomas.viehmann.net/



-- 
To UNSUBSCRIBE, email to debian-boot-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#514016: options for fixing

[Thomas Viehmann]

Hi,

just as an exercise what might be done to fix this:
It would seem that the options are
- not fix it (for now),
- find something in current tar that works (I didn't),
- switch off tar,
- try to do something with tar
  (new upstream - vs. release: ugh - has a "apply to symlink target"
   --transform flag, but that also might need post-processing of
   symlinks at end of bootstrapping),
- try to do some ld-preload(?) trickery for changing the way symlinks
  are read,
- try to find some way to chroot before calling tar, one possibility
  would be copying required files to some temp CWD, make tar look there
  for the libraries, chrooting to the target and calling tar from (the
  unchanged) CWD,
- add chroot option to tar (see patch).

It would seem that in terms of total invasiveness, the last option is
minimal, but it might be unwise to add options without consulting
upstream (however, it once upstream agrees to have an option with a
given spec, one could implement that locally).

Of course, the patch introduces new strings (translator-:( ), but they
only shown with the new option.

But then, this is just an idea, I'm sure you'll come up with a good, and
probably better solution.

Kind regards

T.
-- 
Thomas Viehmann, http://thomas.viehmann.net/
reverted:
diff -u tar-1.20/src/list.c tar-1.20/src/list.c
--- tar-1.20/src/list.c
+++ tar-1.20/src/list.c
@@ -72,6 +72,11 @@
   name_gather ();
 
   open_archive (ACCESS_READ);
+  if (chroot_option) 
+  {
+ sys_chroot(chroot_option);
+  }
+
   do
 {
   prev_status = status;
diff -u tar-1.20/debian/changelog tar-1.20/debian/changelog
--- tar-1.20/debian/changelog
+++ tar-1.20/debian/changelog
@@ -1,3 +1,10 @@
+tar (1.20-1.1) unstable; urgency=low
+
+  * Non-maintainer upload.
+  * Implement --chroot option for extraction.
+
+ -- Thomas Viehmann   Thu, 05 Feb 2009 17:32:49 +0100
+
 tar (1.20-1) unstable; urgency=low
 
   * new upstream version
only in patch2:
unchanged:
--- tar-1.20.orig/src/system.c
+++ tar-1.20/src/system.c
@@ -24,6 +24,11 @@
 
 #if MSDOS
 
+void
+sys_chroot (char* dir)
+{
+}
+
 bool
 sys_get_archive_stat (void)
 {
@@ -120,6 +125,21 @@
 
 static struct stat archive_stat; /* stat block for archive file */
 
+void
+sys_chroot (char* dir)
+{
+  if (chroot(dir) < 0)
+{
+  int e = errno;
+  FATAL_ERROR ((0, e, _("Cannot chroot to %s"), dir));
+}
+  if (chdir("/") < 0)
+{
+  int e = errno;
+  FATAL_ERROR ((0, e, _("Cannot chdir to '/' after chroot to %s"), dir));
+}   
+}
+
 bool
 sys_get_archive_stat (void)
 {
only in patch2:
unchanged:
--- tar-1.20.orig/src/tar.c
+++ tar-1.20/src/tar.c
@@ -252,6 +252,7 @@
   CHECK_DEVICE_OPTION,
   CHECKPOINT_OPTION,
   CHECKPOINT_ACTION_OPTION,
+  CHROOT_OPTION,
   DELAY_DIRECTORY_RESTORE_OPTION,
   HARD_DEREFERENCE_OPTION,
   DELETE_OPTION,
@@ -613,6 +614,8 @@
 
   {"add-file", ARGP_KEY_ARG, N_("FILE"), 0,
N_("add given FILE to the archive (useful if its name starts with a dash)"), GRID+1 },
+  {"chroot", CHROOT_OPTION, N_("DIR"), 0,
+   N_("chroot to directory DIR for extraction"), GRID+1 },
   {"directory", 'C', N_("DIR"), 0,
N_("change to directory DIR"), GRID+1 },
   {"files-from", 'T', N_("FILE"), 0,
@@ -1590,6 +1593,10 @@
 	args->version_control_string = arg;
   break;
 
+case CHROOT_OPTION:
+  chroot_option = arg;
+  break;
+
 case DELAY_DIRECTORY_RESTORE_OPTION:
   delay_directory_restore_option = true;
   break;
only in patch2:
unchanged:
--- tar-1.20.orig/src/common.h
+++ tar-1.20/src/common.h
@@ -141,6 +141,9 @@
 /* Print a message if not all links are dumped */
 GLOBAL int check_links_option;
 
+/* Chroot option value.  */
+GLOBAL char const *chroot_option;
+
 /* Patterns that match file names to be excluded.  */
 GLOBAL struct exclude *excluded;
 
@@ -293,6 +296,7 @@
 
 /* Specified value or pattern.  */
 GLOBAL const char *volume_label_option;
+
 
 /* Other global variables.  */
 
@@ -700,6 +704,7 @@
 
 /* Module system.c */
 
+void sys_chroot (char* dir);
 void sys_detect_dev_null_output (void);
 void sys_save_archive_dev_ino (void);
 void sys_drain_input_pipe (void);
only in patch2:
unchanged:

Bug#497110: partman-dmraid: obsolete component; should not be included in Lenny

[Thomas Viehmann]

Hi,

Frans Pop wrote:
> Please remove partman-dmraid (source and udeb) from Lenny.

> Because of recent changes in libparted, partman and other D-I components 
> that have now migrated to testing this udeb is no longer needed. I'll 
> also file a BR against ftp.d.o requesting removal from unstable.
The unstable removal is done, so IIUIC the package should be on its way
out of testing as well.

Kind regards

T.
-- 
Thomas Viehmann, http://thomas.viehmann.net/



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

OK to lower priority of mpack?

[Thomas Viehmann]

Hi,

#416569 asks for lowering the priority of mpack from standard to optional.
If you don't have any objections, we'd do this now instead of waiting
post-lenny... Comments?

Kind regards

T.
-- 
Thomas Viehmann, http://thomas.viehmann.net/


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#472997: provide all guided recepies using free space instead of full disk

[Thomas Viehmann]

Package: partman-auto
Version: 75
Severity: wishlist

Hi,

thanks for the installer. I just used lenny Beta 1 and it worked well  
for me!
One thing I would like to have had was all "Guided" recipies (crypt,  
lvm) in a variant that only uses the free space instead of the full  
disk.


Kind regards

T.



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Forward: The 'udebn' problem

[Thomas Viehmann]

Jared Rhine wrote:
> Any response would be appreciated.  Is this a bug or not?  Does this
> issue not hit anyone else when building from a fresh CVS?

This is not strictly a bug, but a build-dependence on sed/unstable.
However, I'd say a space instead of \n would do as well in most cases.

Cheers

T.



pgp0.pgp
Description: PGP signature

Re: d-i debcamp in Oldenburg 2003-09-25 to 2003-09-28

[Thomas Viehmann]

Petter Reinholdtsen wrote:
>   Thomas Viehmann - ?
I'm very sorry to have accepted in the meantime another invitation to a summer
camp, so I cannot visit this. :(

Cheers

T.

P.S.: It would have been Bonn, Germany.


pgp0.pgp
Description: PGP signature

Re: Debconf vs. graphical installer

[Thomas Viehmann]

Sebastian Ley wrote:
>>>I like sebastians proposal of having frontends in librarys for every
>>configuration udeb.
> 
> The more I think of it, the more I start to dislike it ;-) Imagine how
> much work it really is to do such a thing. There must be 
> #udebs x #frontends of custom widgets. If a change in the module
> involves a change in the UI, all frontend widgets must be changed. I
> don't know if that is feasable...

If this is the problem, maybe keeping non-graphical installs as is and having a
parallel system for the graphical installer is an option. That would still mean
a lot of duplicated effort, but I guess if the alternatives boil down to an
overdesigned "this handles every case" it's still worth a thought. (Maybe only a
short one, though...)

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: New d-i debcamp in a few months?

[Thomas Viehmann]

Petter Reinholdtsen wrote:
> [Petter Reinholdtsen]
> One initial idea is to arrange it from Monday 2003-10-27 to Friday
> 2003-10-31 at http://www.einschlingen.de/>.  To get this location
> we need to make a quick decision, but I guess we need to give everyone
> some days to get back from debcamp/debconf3 before we can expect an
> answer to the following question:

Don't know if I'm part of the targeted potential participants.
I don't think you should change any plans just because of me (as if you would),
but I probably cannot leave for a whole business week in October (or later).
Also this coincides with Linux World Expo in Frankfurt (2003-10-2[789]).

Cheers

T.



pgp0.pgp
Description: PGP signature

Removal of user mount hack / root requirement for building debianinstaller

[Thomas Viehmann]

Hi.

In

I suggested a patch to make use of syslinux's new capability to write to
unmounted images in order to be able to build as normal user.
A oneliner to allow older sed versions to be used (seperate module names by " "
instead of "\n") was also included.
I've attached an updated patch.

Is this entirely unattractive or just not a priority right now?
If the latter is the case, which would be the most appropriate package to file
wishlist bugs against?

Cheers

T.
--- linux-i386.orig Sun Jul 13 10:44:02 2003
+++ linux-i386  Sun Jul 13 10:44:02 2003
@@ -8,31 +8,12 @@
dd if=/dev/zero [EMAIL PROTECTED] bs=1k count=$(FLOPPY_SIZE)
mkfs.msdos -i deb1 -n 'Debian Installer' -C [EMAIL PROTECTED] 
$(FLOPPY_SIZE)
 
-ifdef USER_MOUNT_HACK
-   ln -sf `pwd`/[EMAIL PROTECTED] $(USER_MOUNT_HACK)
-   mount $(TMP_MNT)
-else
-   mount -t vfat -o loop [EMAIL PROTECTED] $(TMP_MNT)
-endif
-
-   # syslinux is used to make the floppy bootable.
-   if $(foreach NAME,$(KERNELNAME), \
- cp -f $(TEMP)/$(NAME) $(TMP_MNT)/linux) \
-  && cp $(INITRD) $(TMP_MNT)/initrd.gz \
-  && cp boot/i386/syslinux.cfg $(TMP_MNT)/ \
-  && todos $(TMP_MNT)/syslinux.cfg ; \
-   then \
-   umount $(TMP_MNT) ; \
-   true ; \
-   else \
-   umount $(TMP_MNT) ; \
-   false ; \
-   fi
-
-ifdef USER_MOUNT_HACK
-   syslinux $(SYSLINUX_OPTS) $(USER_MOUNT_HACK)
-   rm -f $(USER_MOUNT_HACK)
-else
+   # syslinux is used to make the floppy bootable
+   $(foreach NAME,$(KERNELNAME), \
+ mcopy [EMAIL PROTECTED] $(TEMP)/$(NAME) ::\linux \
+  && mcopy [EMAIL PROTECTED] $(INITRD) ::\initrd.gz \
+  && todos < boot/i386/syslinux.cfg |  \
+   mcopy [EMAIL PROTECTED] - ::\syslinux.cfg)
syslinux $(SYSLINUX_OPTS) [EMAIL PROTECTED]
 endif
 
--- build/Makefile~ Sat Jun 14 17:18:20 2003
+++ build/Makefile  Sat Jun 14 17:20:46 2003
@@ -49,7 +49,7 @@
pkg-lists/base \
pkg-lists/$(TYPE)/common \
`if [ -f pkg-lists/$(TYPE)/$(DEB_HOST_ARCH) ]; then echo 
pkg-lists/$(TYPE)/$(DEB_HOST_ARCH); fi` \
-   | sed -e 's/^\(.*\)$${kernel:Version}\(.*\)$$/$(foreach 
VERSION,$(KERNELIMAGEVERSION),\1$(VERSION)\2\n)/g' \
+   | sed -e 's/^\(.*\)$${kernel:Version}\(.*\)$$/$(foreach 
VERSION,$(KERNELIMAGEVERSION),\1$(VERSION)\2 )/g' \
) $(EXTRAS)
 # Scratch directory.
 BASE_TMP=./tmp/


pgp0.pgp
Description: PGP signature

Re: Debian Installer building (getting rid of user-mount hack andtypo)

[Thomas Viehmann]

Petter Reinholdtsen wrote:
> Where is this patch?  Is it commited?
Sorry, I forgot to attach it. (Luckily only the diff was in /tmp, not the files...)
Thus, I don't think it's commited.

Cheers

T.
--- build/Makefile~ Sat Jun 14 17:18:20 2003
+++ build/Makefile  Sat Jun 14 17:20:46 2003
@@ -49,7 +49,7 @@
pkg-lists/base \
pkg-lists/$(TYPE)/common \
`if [ -f pkg-lists/$(TYPE)/$(DEB_HOST_ARCH) ]; then echo 
pkg-lists/$(TYPE)/$(DEB_HOST_ARCH); fi` \
-   | sed -e 's/^\(.*\)$${kernel:Version}\(.*\)$$/$(foreach 
VERSION,$(KERNELIMAGEVERSION),\1$(VERSION)\2\n)/g' \
+   | sed -e 's/^\(.*\)$${kernel:Version}\(.*\)$$/$(foreach 
VERSION,$(KERNELIMAGEVERSION),\1$(VERSION)\2 )/g' \
) $(EXTRAS)
 # Scratch directory.
 BASE_TMP=./tmp/
--- build/make/arch/linux-i386.orig Sat Jun 14 11:13:58 2003
+++ build/make/arch/linux-i386  Sat Jun 14 17:34:51 2003
@@ -8,33 +8,13 @@
dd if=/dev/zero [EMAIL PROTECTED] bs=1k count=$(FLOPPY_SIZE)
mkfs.msdos -i deb1 -n 'Debian Installer' -C [EMAIL PROTECTED] 
$(FLOPPY_SIZE)
 
-ifdef USER_MOUNT_HACK
-   ln -sf `pwd`/[EMAIL PROTECTED] $(USER_MOUNT_HACK)
-   mount $(TMP_MNT)
-else
-   mount -t vfat -o loop [EMAIL PROTECTED] $(TMP_MNT)
-endif
-
-   # syslinux is used to make the floppy bootable.
-   if $(foreach NAME,$(KERNELNAME), \
- cp -f $(TEMP)/$(NAME) $(TMP_MNT)/linux) \
-  && cp $(INITRD) $(TMP_MNT)/initrd.gz \
-  && cp syslinux.cfg $(TMP_MNT)/ \
-  && todos $(TMP_MNT)/syslinux.cfg ; \
-   then \
-   umount $(TMP_MNT) ; \
-   true ; \
-   else \
-   umount $(TMP_MNT) ; \
-   false ; \
-   fi
-
-ifdef USER_MOUNT_HACK
-   syslinux $(SYSLINUX_OPTS) $(USER_MOUNT_HACK)
-   rm -f $(USER_MOUNT_HACK)
-else
+   # syslinux is used to make the floppy bootable
+   $(foreach NAME,$(KERNELNAME), \
+ mcopy [EMAIL PROTECTED] $(TEMP)/$(NAME) ::\linux \
+  && mcopy [EMAIL PROTECTED] $(INITRD) ::\initrd.gz \
+  && todos < syslinux.cfg |  \
+   mcopy [EMAIL PROTECTED] - ::\syslinux.cfg)
syslinux $(SYSLINUX_OPTS) [EMAIL PROTECTED]
-endif
 
# Finalize the image.
mv [EMAIL PROTECTED] $@
@@ -49,4 +29,3 @@
 cd_image: cd_content
mkisofs -r -J -b `basename $(IMAGE)` -o $(TYPE).iso $(DEST)
mv $(TYPE).iso $(DEST)
-


pgp0.pgp
Description: PGP signature

Debian Installer building (getting rid of user-mount hack and typo)

[Thomas Viehmann]

Hi.

Recently, an updated syslinux got uploaded. It does not insist on mounting
images anymore. I've created a patch eliminating USER_MOUNT_HACK from the
$(IMAGE) target in d-i/build/arch/linux-i386. (But not from the clean target,
this might be added now or later.)
While testing (with "fakeroot make fakeroot make image", or rather with
"TYPE=net FLOPPY_SIZE=2880" added due to disk full errors), I noticed that the
assignment of the UDEBS variable at the top of d-i/build/Makefile has (for the
kernel udebs) a "\n" on the subsitution side of a sed s/.../.../ command, which
does not work with the woody version of sed. While this is clearly a minor to no
bug, I think a space would work as well in that situation.

Possibly, the elimination of USER_MOUNT_HACK should wait until syslinux 2.0.4
hits testing.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Suggestions needed for gtk frontend for cdebconf

[Thomas Viehmann]

Bastian Blank wrote:
>>Against X:
>>- larger than libdirectfb (is this true?)
[Some numbers which may or may not be applicable.]

IMHO really one would have to compare libdirectfb+gtk+2.0-directfb with
X-Server+X-Client lib+Toolkit.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Suggestions needed for gtk frontend for cdebconf

[Thomas Viehmann]

My apologies for writing German.
Thomas Viehmann wrote:
> Sebastian Ley wrote:
> 
> 
>>libdirectfb does not support vt switching, so I searched for a possibility
>>to use a graphical terminal. There are terminal widgets for gtk (vte, zvt)
>>but they require the X target of gtk, so they will only work with an X
>>server.
> 
> 
> 1. zvt ist kleiner (ist ja auch erklärtes Ziel).
1. zvt is smaller (which it aims to be)
> 2. a. In zvt scheint X für Text-Ausgabe benutzt zu werden.
2.a. In zvt the primary direct usage of X seems to be for text output.
>b. Eventuell könnte man das durch directfb-Routinen ersetzen.
  b. Maybe one could replace that by directfb calls.
> Wenn Du immer noch suchst, würde ich mir mal 2b angucken.
If this is still the direction, I'd look at 2b.
(But as Sebastian points out, a mini-X-Server might be preferred.)

Cheers+ Sorry again

T.


pgp0.pgp
Description: PGP signature

Re: Suggestions needed for gtk frontend for cdebconf

[Thomas Viehmann]

Sebastian Ley wrote:

> libdirectfb does not support vt switching, so I searched for a possibility
> to use a graphical terminal. There are terminal widgets for gtk (vte, zvt)
> but they require the X target of gtk, so they will only work with an X
> server.

1. zvt ist kleiner (ist ja auch erklärtes Ziel).
2. a. In zvt scheint X für Text-Ausgabe benutzt zu werden.
   b. Eventuell könnte man das durch directfb-Routinen ersetzen.
Wenn Du immer noch suchst, würde ich mir mal 2b angucken.

Gruß

T.


pgp0.pgp
Description: PGP signature

Re: Building d-i as a package

[Thomas Viehmann]

Alastair McKinstry wrote:
> On Sun, 2003-04-06 at 08:12, Petter Reinholdtsen wrote:
>>And I am fairly sure the buildd is compiling as a normal user and not
>>root.  This will at the moment make it impossible to build the boot
>>floppies.  I suspect this package will fail to build on all platforms
>>until mtools/syslinux will work on an image instead of a loopback
>>mounted device.
> Hmm. More work than I had hoped to have to do. But if it has to be done,
> it'll be done.
Recent upstream syslinux already does support that, I've filed a wishlist bug
against syslinux and use a private version successfully for that. (Cf. my
message from April 3rd in this thread.)

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Woody boot-floppies use flawed kernels

[Thomas Viehmann]

Petter Reinholdtsen wrote:
> [Thomas Viehmann]
> 
>>For installation at least, a local root hole is completely
>>irrelevant. (There is no root password and no users.)
>>The only thing that needs to be ensured is that the installed kernel
>>is not vulnerable. That means
>>- until a new point release is made, stock kernels should be automatically
>>  upgraded via the security.d.o apt-lines,
> 
> 
> This do not work as you expect it.  The kernel used by b-f is copied
> into place on the HD by b-f.  There is no package to upgrade.  No
> kernel package is installed by b-f, and the people with a stock woody
> will have the security problem until they manually install a new and
> improved kernel.

Yes. I see now. Sorry for the misinformation.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Woody boot-floppies use flawed kernels

[Thomas Viehmann]

HI Jonathan,

Jonathan Quick wrote:
>   Is there any intention to release a new version of the Woody boot-floppies
> based on the kernel-image-2.2.25 and kernel-image-2.4.20-1 kernels which 
> include the ptrace security hole fix ( see DSA-270 for example.)  Obviously
> this would require similar patched kernels for all architectures to be
> available too.  Perhaps a critical or grave bug should be filed against
> the boot-floppies & debian-cd to ensure this issue receives attention ?

For installation at least, a local root hole is completely irrelevant. (There is
no root password and no users.)
The only thing that needs to be ensured is that the installed kernel is not
vulnerable. That means
- until a new point release is made, stock kernels should be automatically
  upgraded via the security.d.o apt-lines,
- when a new point release is made, fixed kernels should be offered to install
  on the hard disk.
Unless there is a problem with one of these, I don't think there's much of a
bug, certainly not in boot floppies.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Building d-i as a package

[Thomas Viehmann]

Alastair McKinstry wrote:
> The plan would be that debuild in ./debian-installer/build builds
> debian-installer-$(ARCH)_$(VERSION)_all.deb, that puts the images and
> associated documentation into a place suitable for an archive, 
> eg. /var/www/debian/dists/$(DIST)/main/disks-$(ARCH)/$(VERSION)/..
That would be very cool. Judging from my limited experience with building, the
syslinux should be updated for proper building without root, though.
I have filed a whishlist bug in addition to the debian installer bug.
In the meantime I've built my own 2.02 packages and changed the build/Makefile
$(FLOPPY_IMAGE) target to something along the lines of
# syslinux is used to make the floppy bootable.
mcopy -i$(FLOPPY_IMAGE).new $(KERNEL) ::linux
mcopy -i$(FLOPPY_IMAGE).new $(INITRD) ::initrd.gz
todos < syslinux.cfg > $(TEMP)/syslinux.cfg
mcopy -i$(FLOPPY_IMAGE).new $(TEMP)/syslinux.cfg ::syslinux.cfg

syslinux $(SYSLINUX_OPTS) $(FLOPPY_IMAGE).new
If course, I'll happily post a patch for the whole deal (cleaning out user mount
hack etc.).

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: partitioner_0.01_i386.changes REJECTED

[Thomas Viehmann]

James Troup wrote:
> Debian Installer <[EMAIL PROTECTED]> writes:
>>Rejected: partitioner_0.01_i386.udeb: architecture part of filename
>>  (i386) does not match package architecture in the udeb
>>  (all).
> Wow, good effort.  I never expected anyone to trip _that_
> rejection. ;)

Maybe things like this are an argument for having variants of some
dpkg-dev/debhelper scripts for creating udebs.
The missing priority field issue seemed like this, too. In fact, I gathered that
this was on purpose while I tried to learn by example.
Other example where scripts would help are automatically creating udeb-dev
packages from the udeb library's build tree and having shlib-Dependencies that
don't necessarily count on anna not caring for versions and thus being
statisfied with versioned dependencies easily.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: Bug#179000: Handling kernel modules of different kernel versionsin anna & d-i

[Thomas Viehmann]

Hi.

Martin Sjögren wrote:
> Any comments on this?
IMHO the *.menutest scripts seem to be suited ideally for that.

Cheers

T.


pgp0.pgp
Description: PGP signature

Sorry (Re: lib*.so.* -Frage)

[Thomas Viehmann]

Sorry. This was a mispost.
Cheers
T.




pgp0.pgp
Description: PGP signature

Re: lib*.so.* -Frage

[Thomas Viehmann]

Hallo.

Frederik wrote:
Zu den *.so*-Dateien:
Solange die Debian-Pakete gleich heißen, werden die Pakete und damit auch die
darin enthaltenen Dateien (in denen, die nicht *-dev heißen, so.x.y.z und so.x)
ersetzt. Die Idee dabei ist, daß alle so.x.*.* für andere Programme kompatibel
sind ("die ABIs sind kompatibel"). Insofern sollte dieses ersetzen auch keine
Effekte haben. (Bei unstable weiß man aber, nomen est omen, nie, obwohl es, je
nach Maintainer, eher selten ist, daß etwas gar nicht mehr geht.)
Wenn sich nun die so-Version (das x in so.x.y.z) ändert soll sich auch der
Paketname ändern (der sollte nämlich lib*x sein).


> wenn ich jetzt weiss bzw. vermute dass, package sonso schuldig ist,
> dachte ich mit apt-get install package=[Versionsnummer] die ältere Version 
> wiederzubekommen!? 
Wenn sie in den Apt-Quellen, die Du hast, ersetzt worden sind, nicht. Du kannst
Dir eventuell damit behelfen, daß Du mit dpkg --remove und dpkg --install und
dem Apt-Cache (/var/cache/apt/archives/) arbeitest.
Oder Du trägst etwas wie
deb http://snapshot.debian.net/archive pool pkgname
in deine Sources list ein, dann bekommst Du alle (neueren) alten Versionen.
*Eigentlich* sollte das aber nicht so oft nötig sein, zumal, wenn Du stable benutzt.

[Missing symbol]
Eventuell solltest Du mal sehen, ob Du da Packages mit neu kompilieren mußt.
(evtl. mit gcc-3.2). Vielleicht ist auch ein selbstkompiliertes qt keine so
tolle Idee, wenn Du (offenbar?) libxft aus unstable benutzt(?). (D.h. wenn Du es
neu kompilieren mußtest, weil das aus Unstable nicht funktionierte, hast Du
vielleicht ein anderes Problem.)

Gruß

T.



pgp0.pgp
Description: PGP signature

Re: trying to install but can't recognize the disk....

[Thomas Viehmann]

Walter Tautz wrote:
> the system came preinstalled
> with redhat 8.0 and I do have a copy of the kernel-config file
> for the running kernel so if I knew how to build a different install disk...
If you want to equip a boot disk with another kernel, Chapter 10.3 of the Debian
Installation Guide [0] may help you.
Probably you will have to take care of installing that kernel on your hard drive
 manually, but this always works for me.

Cheers

T.

0.
http://www.debian.org/releases/stable/i386/ch-boot-floppy-techinfo.en.html#s-rescue-replace-kernel

P.S.: Strictly speaking, the correct place to ask would be debian-user, as
debian-boot is intended for development. Most people won't mind that much, but
some get offended more easily.


pgp0.pgp
Description: PGP signature

Re: 'net' boot floppy is full and unable to build.

[Thomas Viehmann]

Hi.

At least on my recent builds, there is libbogl included, which probably should
be removed along with the bogl frontend. (Approx. 100k in tree, removing it
solves the problem for me).

Quoting Martin Sjögren:
> This is because the last upload of cdebconf was broken and included the
> bogl frontend as well as the text frontend. A new cdebconf is uploaded,
> but it's NEW, so it'll take some time.
>
> Oh, and don't worry about the library missing, the broken cdebconf
> probably doens't fit on a net floppy anyway. :(

I wouldn't know anything, though.

Cheers

T.



pgp0.pgp
Description: PGP signature

Re: Prompt: 1 - 19, default=Detect a keyboard and select layout>

[Thomas Viehmann]

> kbd-chooser's postinst exited with status 256
...
> There's no message anywhere I can see to tell me what's wrong.
> 
> FWIW manual selection does not work either.
> FWIW I have bootkbd=qwerty/us on the kernel commandline.
kbd-chooser is failing? (Don't know why though.)

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: [OT] installation documentation

[Thomas Viehmann]

Chris Tillman wrote:
>>Might be nice to add sth like :
>>"To create a swap partition, you need to create a partition with the 
>>name "swap" AND the type "Apple_UNIX_SRV2".
> I don't really read French; but this doesn't do it?
No. It does not mention the type Apple_UNIX_SRV2. (I guess that's why Fabrice
capitalized the AND.) Neither does the English version which is indeed
faithfully translated.
Also the reference to the English-only mac-fdisk Tutorial is less nice from a
non-English manual if it's really required. The worst is the "** man page not
found **" I get from the link to the mac-fdisk manpage.

> Apple_UNIX_SRV2 is the type of partition that mac-fdisk creates by
> default, so it's not necessary to know.
I don't know the technical side of this issue, but if mac-fdisk asked for a type
and allows you to change it and if it was necessary that it'd be
Apple_UNIX_SRV2, it would be misleading to state "swap partition is identified
on Mac type disks by its name", because it suggests that the type doesn't matter.

Cheers

T.


pgp0.pgp
Description: PGP signature

Re: rfc: A guide for packaging library udebs

[Thomas Viehmann]

Cardenas wrote:
> On Mon, Mar 03, 2003 at 11:57:04PM +0100, Thomas Viehmann wrote:
[...]
>>what would go wrong unless one does the right thing.) In this particular case,
>>it might be interesting to know what the advantage of a subdirectory for each
>>udeb library over a /usr/lib/udeblib is.
> take the example of glib. if you have special compile options for
> glib in d-i, then you create a library with a different soname, and
> different abi, but the same filename as the standard glib thats on
> your machine. in order to link a d-i module against that glib, you
> would have to have the glib-udeb version of the library on your
> system. you wouldn't want to remove your standard glib, right? so, to
> have both installed, you should use a subdir of /usr/lib. 
Yes, I see why they cannot be in /usr/lib, but what's the rationale to seperate
the udeb libraries amongst each other rather than throw all d-i libs in one
directory? (Yes, I admit that I don't need to know for doing anything, but it'll
still is a question that made me think.)

Cheers

Thomas


pgp0.pgp
Description: PGP signature

Re: rfc: A guide for packaging library udebs

[Thomas Viehmann]

Hi.

I don't know if how general the audience you're aiming at is. I'm not an expert,
but rather a casual user who recently built a library udeb for his private
purposes (it was slang1a for nano, I used libpango and others as examples and it
worked in the end).
Most of the problems I ran into are commented in you're guide, however I believe
that some questions might be worth being more clear about.

Sebastian Ley wrote:
> To be able to use dh_shlibdeps anyway each library module should add a
> "Provides:" line, providing the name of the corresponding deb package
> of the library. I.e. libc6-udeb should provide libc6.
I do have the impression that most shlibs files are versioned dependencies. It
is my understanding that the debian-installer simply ignores them so that a
simple provides is sufficient. However, this is a difference to conventional deb
packages, for which it is (due to lack of versioned provides) currently
impossible to satisfy such a versioned dependency (AFAIK).

[...]
> Since modules should not be installed on normal systems there is need
> for a deb package providing this library as well as the necessary
> devekopment link. The package should be named libfoo-udeb-dev and
> declare a dependancy on libfoo-dev. The only files libfoo-udeb-dev
> should provide is the ABI-changed library and the necessary
> symlinks. Since the development link is identically to the development
> link in libfoo-dev, the library and the links should be installed in a
> subdirectory of /usr/lib, e.g. /usr/lib/libfoo-udeb.
If you want to make this a standard, you should IMHO be more precise about the
subdirectory: First it should probably not be "e.g.", but rather "the
subdirectory should be named..." and then you might want to specify the exact
derivation of the name. (Is it the derived from the library soname, the package
name, should the soversion be used? Sometimes, those the package name and the
library name differ, e.g. slang1a).
Also, one of the things I like about the Junichi Uekawa's libpkg-guide is that
it tries to explain the rationale. (At least I myself try to learn by imagining
what would go wrong unless one does the right thing.) In this particular case,
it might be interesting to know what the advantage of a subdirectory for each
udeb library over a /usr/lib/udeblib is.
Also, what happens with api incompatibilities? Is it entirely impossible to have
different include libraries for the "real" lib debs and the udeb versions?

Your guide has been very interesting reading. Thank you.

Cheers

T.


pgp0.pgp
Description: PGP signature

Compiling udebs and creating cdrom images

[Thomas Viehmann]

Hi.

As you probably have noticed, I'm trying out debian-installer in funny ways.
This time I wanted to do was compile a cdrom image, however I failed and have
some questions. If you can spare a little time, I'd really appreciate you help.

- Are the pkglists in cvs up to date?
  They seem to include cdrom-checker which is uninstallable because it doesn't
  depend on cdebconf-udeb but on cdebconf.) I noticed that the autobuild logs
  (which unfortunately don't have the statistics in the end) seem to include
  a smaller set of packages.
  (Of course, if you don't have the time to explain, but have scripts for
   doing things, (a pointer to) the scripts would also be appreciated.)
- How are udebs properly compiled? While creating a patch to get a
  slang udeb (for nano) I noticed that a rather plain dh_shlibdeps is used in
  libpango, however I'm at a loss at how to make it use the -udeb dependencies
  rather than the full ones (e.g. libc6-udeb instead of libc6 (>>...)).
- Are the cd-images still created using debian-cd? (Not that I did find any
  evidence why they shoudln't, but I'm just courious.)

Thanks in advance

Thomas

BTW: If you have (semi-) useful tasks to hand out to newbies, I'd be glad to see
if I can return something for the time you spend on helping me with my
questions. I know my skills and knowledge are rather limited, but they always
tell you to see about helping debian-installer and I guess everything I try with
my own motivation either causes more work for the developers than helping them
(e.g. the sudo in the make floppy_images)) or is plainly useless (as with the
slang udebs - it is my understanding that the request for properly renaming pic
files obsoletes the want for udebs, but I may be mistaken).




pgp0.pgp
Description: PGP signature

patch to create slang1a-utf8-udeb

[Thomas Viehmann]

Hi.
I was sort of annoyed by nano not working in debian-installer.
The result is the attached patch, which changes this.
As I'm just a random user of nano in debian-installer initrd, you probably want
to wait what someone of the debian-installer team has to say about this.
Especially, I don't know about the hardcoded dependency on libc6 and whether bug
#182042 obsoletes this bug.

Cheers

Thomas
diff -urN x/slang-1.4.5/debian/control slang-1.4.5/debian/control
--- x/slang-1.4.5/debian/controlFri Feb 28 00:43:27 2003
+++ slang-1.4.5/debian/control  Fri Feb 28 00:28:11 2003
@@ -112,3 +112,15 @@
  on custom installation floppies and in embedded systems. Unless you're
  making one of those, you won't need this package.
  This packages has wide character support.
+
+Package: slang1a-utf8-udeb
+Section: debian-installer
+Priority: optional
+Architecture: any
+Depends: libc6-udeb
+Description: S-Lang library with utf8 support
+ This is a udeb, or a microdeb, of the S-Lang library with wide charater
+ support. As such it is the installer counterpart of slang1a-utf8.
+ You only need this package to support applications needing S-Lang during
+ the Debian installation process time and probably don't need to select it
+ manually for installation.
diff -urN x/slang-1.4.5/debian/rules slang-1.4.5/debian/rules
--- x/slang-1.4.5/debian/rules  Fri Feb 28 00:43:27 2003
+++ slang-1.4.5/debian/rulesFri Feb 28 00:45:42 2003
@@ -1,4 +1,5 @@
 #!/usr/bin/make -f
+# udeb adaptation by Thomas Viehmann
 # Made with the aid of dh_make, by Craig Small
 # Sample debian/rules that uses debhelper. GNU copyright 1997 by Joey Hess.
 # This version is for a hypothetical package that builds an
@@ -19,6 +20,10 @@
 
 LIBSLANG=slang1
 LIBSLANG_UTF8=slang1a-utf8
+LIBSLANG_UTF8_UDEB=$(LIBSLANG_UTF8)-udeb
+
+DEBVERSION=$(shell dpkg-parsechangelog | grep '^Version: ' | sed -e 's/^Version: //')
+UDEBNAME=$(LIBSLANG_UTF8_UDEB)_$(DEBVERSION)_$(shell dpkg-architecture 
-qDEB_BUILD_GNU_CPU).udeb
 
 build: 
true;
@@ -105,10 +110,12 @@
(sed 's/@DEBIANUTF8ERRORCHECK@/#ifndef/' < debian/slang.h.extra.in; cat 
src/slang.h ) > `pwd`/debian/slang1-utf8-dev/usr/include/slang.h
chmod 644 `pwd`/debian/slang1-utf8-dev/usr/include/slang.h
 
+   cp debian/slang1-utf8-dev/usr/lib/libslang.so.* 
debian/$(LIBSLANG_UTF8_UDEB)/lib/libslang.so.$(SOMAJOR)-UTF8.$(SOMINOR)
mv debian/slang1-utf8-dev/usr/lib/libslang.so.* 
debian/slang1a-utf8/lib/libslang.so.$(SOMAJOR)-UTF8.$(SOMINOR)
 
# The ldconfig symlink to make the library work ASAP.  This is not really 
required.
cd debian/slang1a-utf8/lib ; ln -sf libslang.so.$(SOMAJOR)-UTF8.$(SOMINOR) 
libslang.so.$(SOMAJOR)-UTF8
+   cd debian/slang1a-utf8-udeb/lib ; ln -sf 
libslang.so.$(SOMAJOR)-UTF8.$(SOMINOR) libslang.so.$(SOMAJOR)-UTF8
 
# Correct the .so link for slang1-utf8-dev library
cd debian/slang1-utf8-dev/usr/lib ; ln -sf 
/lib/libslang.so.$(SOMAJOR)-UTF8.$(SOMINOR) libslang.so
@@ -120,8 +127,8 @@
 binary-arch: binary-nonutf8 binary-utf8
dh_testdir -a
dh_testroot -a
-   dh_installdocs -a
-   dh_installexamples -a
+   dh_installdocs -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_installexamples -a -N$(LIBSLANG_UTF8_UDEB)
 
cd debian/slang1-dev/usr/share/doc/slang1-dev/examples/ ; \
mv demo/* . ; \
@@ -131,10 +138,10 @@
mv Makefile.simple Makefile ; \
rm -f config.status config.log configure configure.in
 
-   dh_installmenu -a
-   dh_installcron -a
-   dh_installmanpages -a
-   dh_installchangelogs -a changes.txt
+   dh_installmenu -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_installcron -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_installmanpages -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_installchangelogs -a changes.txt -N$(LIBSLANG_UTF8_UDEB)
dh_strip -a
dh_compress -a
 
@@ -148,10 +155,15 @@
dh_makeshlibs -p$(LIBSLANG_UTF8) -V "${LIBSLANG_UTF8} (>> 1.4.4-7.1)"
dh_installdeb -a
dh_shlibdeps -a
-   dh_gencontrol 
-
-   dh_md5sums -a
-   dh_builddeb -a
+   dh_gencontrol -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_md5sums -a -N$(LIBSLANG_UTF8_UDEB)
+   dh_builddeb -a -N$(LIBSLANG_UTF8_UDEB)
+
+   # don't know how to do shlibdeps
+   # dh_shlibdeps
+   dh_gencontrol -a -p$(LIBSLANG_UTF8_UDEB) -- -fdebian/files~
+   dpkg-distaddfile $(UDEBNAME) debian-installer optional
+   dh_builddeb -p$(LIBSLANG_UTF8_UDEB) --filename=$(UDEBNAME)
 
 source diff:  
@echo >&2 'source and diff are obsolete - use dpkg-source -b'; false
diff -urN x/slang-1.4.5/debian/slang1a-utf8-udeb.dirs 
slang-1.4.5/debian/slang1a-utf8-udeb.dirs
--- x/slang-1.4.5/debian/slang1a-utf8-udeb.dirs Thu Jan  1 01:00:00 1970
+++ slang-1.4.5/debian/slang1a-utf8-udeb.dirs   Thu Feb 27 23:18:48 2003
@@ -0,0 +1 @@
+lib


pgp0.pgp
Description: PGP signature

Bug#182357: choose-mirror fails when called for the second time

[Thomas Viehmann]

retitle 182357 choose-mirror fails when called for the second time


pgp0.pgp
Description: PGP signature

Bug#182349: anna: Should reload packages file if something went wrong last time

[Thomas Viehmann]

Martin Sjögren wrote:
>>One gets invalid packages files every so often (e.g. during mirror update).
>>However, probably whishlist is better.
> Well, even if the Packages files aren't verified for integrity right
> now, they will be as soon as Release file support is added (which is
> waiting for the archive to adapt the Release file :P).
Then, if it won't let me redownload the release file when something doesn't
match the release file, I shall file another bug.
(d-i seems to be very strict with me when it comes to order of things...)

> Also, how will it help to get a new Packages file right again, won't it
> be just as broken?
The order in which the mirrors take actions doesn't seem to be particulary
predictable. So really, if something in the chain
Release.gpg->Release->Packages->.*\.u?debs fails, one should start from the
beginning, presently this is the Packages file.

The concrete incident that caused me to file the bug was somewhat artificial,
however, I see things like this from time to time when updating my mirror.

Cheers

Thomas


pgp0.pgp
Description: PGP signature

Bug#182349: anna: Should reload packages file if something went wrong last time

[Thomas Viehmann]

Matt Kraai wrote:
> On Mon, Feb 24, 2003 at 10:13:20PM +0100, Thomas Viehmann wrote:
> The Packages file lists packages that have not been mirrored yet?  If
> so, how would downloading it again help?
If the Packages gets updated after the packages.
The downloaded Packages file can list packages that have been replaced by newer
versions, the old ones ceasing to be available.
Then a reload (after the Packages file has been adjusted) helps.

Cheers

Thomas



pgp0.pgp
Description: PGP signature

Bug#182357: choose-mirror: Error message for failure

[Thomas Viehmann]

Package: choose-mirror
Version: N/A; reported 2003-02-24
Severity: Normal

I get this

/home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16 (fronten
d_add): /home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16
(frontend_add): Assertion failed: q->next == NULL
choose-mirror's postinst exited with status 36096

when I'm trying to revise my choice of mirror.
Below is something like a log / instructions how to get there.
(BTW: Is there a log? Oh, well cut and paste from uml is good enough.)

Cheers

Thomas

  2. Configure the network via DHCP
Prompt: 1 - 6> 2
For example, this question is of medium priority, and if your priority were
already 'high' or 'critical', you wouldn't see this question.
  4. low
Prompt: 1 - 4> 4
...
  4. Choose mirror to install from
Prompt: 1 - 6> 4
Download files using what protocol?
  1. http (default)
Prompt: 1 - 2> 1
Choose mirror
 52. enter information manually
Prompt: 1 - 52> 52
Then: 192.168.100.1 /debuan
  4. Choose mirror to install from
Prompt: 1 - 6> 4
Download files using what protocol?
  1. http (default)
Prompt: 1 - 2> 1
Choose mirror
 52. enter information manually (default)
Prompt: 1 - 52> 52
/home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16 (fronten
d_add): /home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16
(frontend_add): Assertion failed: q->next == NULL
choose-mirror's postinst exited with status 36096
installer[42]: Setting main menu question priority to critical

Debian Installer Main Menu



pgp0.pgp
Description: PGP signature

Bug#182349: anna: Should reload packages file if something went wrong last time

[Thomas Viehmann]

Martin Sjögren wrote:
> Just correcting the speeling :)
Please excuse my lack of attention.

> First I thought you meant the Packages files that are downloaded and
> parsed, but then I guess you meant the actual packages? :)
No. I meant the Packages file.

> PS: Why maintonly?
As both Matt and you noted, it's best filed as a whishlist bug. I had a normal
first, then remembered to correct the To for whishlist, but forgot to actually
change the severity pseudo header. (Filing bugs while cooking...)

Cheers

Thomas




pgp0.pgp
Description: PGP signature

Bug#182349: anna: Should reload packages file if something went wrong last time

[Thomas Viehmann]

[Sorry, forgot the bug cc in my first post.]

Matt Kraai wrote:
[anna reloading package files]
> What problem this fix?
One gets invalid packages files every so often (e.g. during mirror update).
However, probably whishlist is better.

Cheers

Thomas



pgp0.pgp
Description: PGP signature

Bug#180032: choose-mirror: Error message for failure

[Thomas Viehmann]

Hi.

Martin Sjögren wrote:
> I don't see how these bug reports are related at all. Isn't this more of
> a cdebconf issue? It fails and everything goes boom. I think you should
> report this on cdebconf instead.
Like the submitter, I had a problem with the download mirror and tried to change it.
>From the original bugreport I figured that he had got back to the main menu.
Also, like him and you, I believe that this is probably a cdebconf problem.

So to me it looked as it was the same bug, that's why I didn't file a new one.
(I can do so, though, if you prefer.) It probably needs to be reassigned.
However, I don't think I'm qualified to reassign this bug. Sorry for not
pointing this out more clearly.

Cheers

Thomas


pgp0.pgp
Description: PGP signature

Bug#182349: anna: Should reload packages file if something went wrong last time

[Thomas Viehmann]

Package: anna
Version: N/A; reported 2003-02-24
Severity: whishlist

I think it would often be preferable if anna (i.e. menu option Load installer
modules) would reload package files if called for the second time.
(However, please feel free to close this bug if you disagree.)

Cheers

Thomas


pgp0.pgp
Description: PGP signature

Bug#180032: choose-mirror: Error message for failure

[Thomas Viehmann]

Followup-For: Bug #180032
Package: choose-mirror
Version: N/A; reported 2003-02-24

Maybe it's nice to see the actual error message. I get the following.
After that, the mirror's name is the empty string.

Cheers

Thomas

...
 49. Turkey
 50. Ukraine
 51. United States
 52. enter information manually (default)
Prompt: 1 - 52> 52
/home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16
(frontend_add):
/home/martin/projects/debian-installer/tools/cdebconf/src/frontend.c:16
(frontend_add): Assertion failed: q->next == NULL
choose-mirror's postinst exited with status 36096
installer[42]: Setting main menu question priority to critical

Debian Installer Main Menu


pgp0.pgp
Description: PGP signature

Bug#182344: di-utils-fake-mkfs: Dependency on partitioned-harddrives is probably incorrect

[Thomas Viehmann]

Package: di-utils-fake-mkfs
Version: N/A; reported 2003-02-24
Severity: minor

Perhaps di-utils-fake-mkfs should provide partitioned-harddrives or something
like that.Otherwise I will get what's shown below, which looks wrong to me.
When I've make filesystems already, I've probably also partitioned the hard drive...
(Similar applies to other fakes.)

Cheers

Thomas

 12. Mount a partition
...
Prompt: 1 - 18> 12
Debian Installer Main Menu

Choose a configuration step
This configuration step requires one or more configuration steps that have not
yet been performed. Please select a configuration step to continue.
  1. Automatically Partition Hard Drives (unsafe) (default)
  2. I have already created file systems
  3. Create a file system
Prompt: 1 - 3> 2
Debian Installer Main Menu

Choose a configuration step
This configuration step requires one or more configuration steps that have not
yet been performed. Please select a configuration step to continue.
  1. Automatically Partition Hard Drives (unsafe) (default)
  2. I have already partitioned the hard drive
  3. Partition a hard drive
Prompt: 1 - 3> 2

-- System Information
Debian Release: 3.0
Architecture: i386
Kernel: Linux hardy 2.4.19-tv #1 Thu Nov 7 12:12:27 CET 2002 i686
Locale: LANG=C, LC_CTYPE=en_US
[EMAIL PROTECTED]:/semitmp/debian-installer/repository$



pgp0.pgp
Description: PGP signature

Re: Running debian-installer within bochs and uml

[Thomas Viehmann]

Matt Kraai wrote:
> On Mon, Feb 24, 2003 at 03:21:12PM +0100, Thomas Viehmann wrote:
> Would you please file a bug report for each problem against the
> appropriate package?  You can assign them to the install
> pseudo-package if you don't know which package is responsible?

I'd really prefer doing some more research on most of this. (E.g. on you
screenshots, there is a en_US locale option.)
Also, there are things that are more design decisions than bugs. For example, no
parameter prompt at default is not a problem when there is an option to increase
the number of questions asked. On the other hand, that option is not needed when
there is an automatic escalation per configuration question. (I think the latter
has been discussed at least for individual configuration questions on the list.)

On the other hand, if you prefer starting this discussion with a bug report,
I'll gladly get busy filing them.

Cheers

Thomas


pgp0.pgp
Description: PGP signature

Running debian-installer within bochs and uml

[Thomas Viehmann]

Hi.

Because I don't have a cdrw burner nor a desire to reboot my computer with the
debian-installer mirror for installing, I've been trying to run debian-installer
within a bochs simulation session and uml.
Bochs is somewhat slow, but the alternative plex86 doesn't seem to support
networking. user-mode-linux doesn't support the mapping block devices as much
AFAIK, so one has to use the ubd devices.

I've made the following observations with bochs, some of which may be bugs (I
didn't check wither they're already reported or were resolved since I've
encountered it, though. I can do this if it's interesting at all.), others of
which will certainly not be bugs, but might be helpful to others who want to try
out debian-installer with bochs.

Then I've tried uml. For that I extended the call in the build/ makefile to the
following:
linux initrd=$(INITRD) root=/dev/rd/0 ramdisk_size=8192 \
   con=fd:0,fd:1 devfs=mount eth0=tuntap,,fe:fd:0:0:0:1,192.168.100.1 \
   ubd0=thedisk.img
thedisk.img being initialized with zeros.
I also found it useful to add a call of "reset" after the linux call.
I can get to the point where one has to create filesystems and mount partitions
easily, but the debian-installer create fs immediately exits with messages as
shown below.
Creating or creating and mounting the /dev/ubd/0 by hand didn't help me, either,
as I still could not continue the installation of the base system.
If noone else is looking into this, I might try to fiddle with the corresponding
udebs, I'd really like to be able to do some debian-installing without having a
spare computer. (Is there anyone currently using d-i on uml? After all, the
Makefile target is there.)
One of the more important things I was missing was an option to readjust the
priorty threshold for questions.

Cheers

Thomas

Some things I did that might be relevant for others testing d-i with bochs:
In the Bochs configuration:
- enable ne2k, floppy image (might use ro), create tun setup script (you need
  sudo for ifconfig)
- Default disk size isn't good for autopartkit.
- Non-English keyboard (mine is German) may have problems with special chars
  ("/","="). '/' is on the numeric keypad, but "=" isn't. I got around the "="
  issue by enabling x11-pc-us.map (us is on purpose) and using the paste option.
  Then, of course, the keyboard layout is funny once a keyboard layout is
  selected in the debian-installer.
- In d-i: ne module needs a param io=0x280, so you need to select questions of
  "low" priority

Some things with debian-installer itself:
- Observations: =,/ for German keyboards
  -> should keyboard table be loaded before network init?
- Messages=Medium: ne doesn't prompt for parameters, not installable
  should prompt me the second time around (similar to tfh's issue with
  autodetection)
- Along the same deal: Invalid installation source: users should be abler to
  eenter server instead of just retreiver when loading fails
- Download percentage is nan% (Maybe this is the lack of a FPU.)
- Locale has the choices (nb|nn|si)_NO|lv_LV ?
- Install base system doesn't succeed:
  [...]
  P: 46 47 Extracting packages
  I: Extracting /var/cache/apt/archives/whiptail_0.50.17-9.6_i386.deb...
  P: 47 47 Extracting packages
  I: Extracting /var/cache/apt/archives/mbr_1.1.5-1_i386.deb...
  P: 1 40 Installing base system
  I: Installing core packages...
  P: 2 40
  base-installer's postinst exited with status 256
  installer[46]: Setting main menu question priority to critical
  Debian Installer Main Menu
  [...]
- It stalled at the installing a kernel step.
  Don't really know what caused this.



 11. Create a file system
Prompt: 1 - 18> 11
Debian Installer Main Menu

Choose a configuration step
This configuration step requires one or more configuration steps that have not
yet been performed. Please select a configuration step to continue.
  1. Automatically Partition Hard Drives (unsafe) (default)
  2. I have already partitioned the hard drive
  3. Partition a hard drive
Prompt: 1 - 3> 2
Warning: start=2 - this looks like a partition rather than
the entire disk. Using fdisk on it is probably meaningless.
[Use the --force option if you really want this]
di-utils-mkfs's postinst exited with status 5120
installer[42]: Setting main menu question priority to critical
Debian Installer Main Menu

Similar for Mount a Partition
or after
# mkdir /target
# mount /dev/ubd/0 /target -t ext2
Prompt: 1 - 18> 13
Debian Installer Main Menu

Choose a configuration step
This configuration step requires one or more configuration steps that have not
yet been performed. Please select a configuration step to continue.
  1. Automatically Partition Hard Drives (unsafe) (default)
  2. I have already mounted the partitions
  3. Mount a partition
Prompt: 1 - 3> 2
cp: /target/etc/fstab: No such file or directory
base-installer's postinst exited with status 256
installer[42]: Setting main menu question priority to critical
Debian Installer Main Menu




pgp0.pgp
Des

Re: Building floppy images without having to mount the image

[Thomas Viehmann]

Bernhard R. Link wrote:
> Perhaps someone knows another tools. Syslinux' main advantage is
> in my eyes that it exists and works. I think it was neighter made for
> this usage (but for beeing able to create such disks from DOS, when
> I didn't misread anything), nor optimal for the usage. (Do we really
> need a FAT-Table on something that is write-once?)
Actually, it's a neat feature that the boot-disk isn't necessarily write once.
For example, the debian installation manual, chapter 10 [0], explains how to
replace the kernel on a boot disk. This is a good thing to recommend to people
having unsupported hardware.

Additionally, syslinux main advantages are that it exists and works and (modulo
the mounting wished to be optional in #180660) is as easy as it gets.
(Amongst syslinux main advantages are such elements as...)

Cheers

Thomas

0.  




msg25857/pgp0.pgp
Description: PGP signature

Re: Building floppy images without having to mount the image

[Thomas Viehmann]

Petter Reinholdtsen wrote:
> [Thomas Viehmann]
> Yes, please.  I really would like to build d-i floppies without being
> root. :)
Matt Kraai noted that the config-stuff isn't needed with the sid mtools (for
them, -i option can be used to specify an image, the no vfat isn't necessary and
but could be passed without configfile).
He filed bug #180660 against syslinux. We agreed later that mcopy'ing the
ldloader.sys directly might be better than having syslinux even try to mount,
and the bugreport was ammended.

Cheers

Thomas



msg25849/pgp0.pgp
Description: PGP signature

Re: Building floppy images without having to mount the image

[Thomas Viehmann]

Matt Kraai wrote:
> On Tue, Feb 11, 2003 at 12:29:24AM +0100, Thomas Viehmann wrote:
> Yes.  I'd prefer to teach syslinux to use mtools, though, rather
> than use ldlinux.sys and ldlinux.bss directly.
Well, probably one would need better mtools before that.
Also I always wondered what /usr/bin/syslinux (i.e. only the program) actually
does. It turns out that apart from checking that it's a dos boot sector it's
overwriting and doing heaps of race avoiding with the mounting of the image, it
copies the ldlinux.sys, adjust the attrib's and copy the right parts of the boot
sector.
In any case, for mtools being of any use to things like this, one should be able
to override the config in a more controlled way. Maybe I'll file a whishlist bug
against it.

Cheers

Thomas



msg25828/pgp0.pgp
Description: PGP signature

Building floppy images without having to mount the image

[Thomas Viehmann]

Hi again.

Maybe this is more productive than the last question.
Is avoiding to be root or modify fstab in order to do make floppy_images
something you could care about?
If it is, I have created together the following patch that does just that.
Note that you need ldlinux.sys and ldlinux.bss from the syslinux build tree
(unfortunately those are compiled into the syslinux binaries, but if you wanted
non-root-buildable floppies, it might worth trying). It also need mtools.
Obviously having the HOME=. for mtools and the dd's to get the bootsector (or
rather the right bits) in isn't too pretty, but it's the best I came up with
without using other packages or even patching mtools.
Of course the whole idea might be rubbish, too, but I just liked the idea of
having an alternative to messing with my fstab or sudo'ing.

Regards

Thomas

--- debian-installer/build/Makefile.origTue Feb 11 00:16:36 2003
+++ debian-installer/build/Makefile Tue Feb 11 00:19:36 2003
@@ -77,6 +77,9 @@
 # or just something extra on a floppy.
 #EXTRAFILES=/usr/bin/strace
 
+# Use mtools to build floppies to avoid having to mount stuff
+PREFER_MTOOLS="yes"
+
 # set DEBUG to y if you want to get the source for and compile 
 # debug versions of the needed udebs
 DEBUG=n
@@ -437,6 +440,7 @@
dd if=/dev/zero of=$(FLOPPY_IMAGE).new bs=1k count=$(FLOPPY_SIZE)
mkfs.msdos -i deb1 -n 'Debian Installer' -C $(FLOPPY_IMAGE).new 
$(FLOPPY_SIZE)
 
+ifndef PREFER_MTOOLS
 ifdef USER_MOUNT_HACK
ln -sf `pwd`/$(FLOPPY_IMAGE).new $(USER_MOUNT_HACK)
mount $(TMP_MNT)
@@ -457,6 +461,21 @@
rm -f $(USER_MOUNT_HACK)
 else
syslinux $(SYSLINUX_OPTS) $(FLOPPY_IMAGE).new
+endif
+else# ifndef PREFER_MTOOLS
+   echo drive a: file=\"$(FLOPPY_IMAGE).new\" exclusive > .mtoolsrc
+   echo MTOOLS_NO_VFAT=1 >> .mtoolsrc
+   HOME=. mcopy $(KERNEL) A:/linux
+   HOME=. mcopy $(INITRD) A:/initrd.gz
+   cp syslinux.cfg $(TEMP)/syslinux.cfg.tmp
+   todos $(TEMP)/syslinux.cfg.tmp
+   HOME=. mcopy $(TEMP)/syslinux.cfg.tmp A:/syslinux.cfg
+   rm $(TEMP)/syslinux.cfg.tmp
+   HOME=. mcopy ldlinux.sys a:\LDLINUX.SYS
+   HOME=. mattrib +r a:\LDLINUX.SYS
+   dd if=ldlinux.bss of=$(FLOPPY_IMAGE).new bs=11 count=1 conv=notrunc
+   dd if=ldlinux.bss of=$(FLOPPY_IMAGE).new bs=62 skip=1 seek=1 conv=notrunc
+   rm .mtoolsrc
 endif
 
# Finalize the image.



msg25821/pgp0.pgp
Description: PGP signature

Building install floppies

[Thomas Viehmann]

Hello.

Sorry if this is just noise, but I tried to build install disks and got the
below after "make build". The result in "make floppy_image" is what can be
predicted from the output.
I have read all the docs in doc/ and the build/README, but will confess that
I've more or less have just added the builddeps from unstable and try to use
fakeroot for make build and su(do) only for make floppy_image.
Is this just the current state of affairs or is there something I'm doing wrong?
(If it's too complicated to explain and you don't have time ATM, that's OK, too,
but I'd love learn about d-i (and in my book this means building it, too.))

Cheers+Thanks

Thomas

System stats

Installed udebs: di-utils-shell rootskel anna main-menu cdebconf-udeb udpkg
busybox-udeb dash-udeb kernel-image-2.4.19-386-udeb choose-mirror net-retriever
netcfg-static netcfg-dhcp pump-udeb modutils-basic ethdetect
nic-modules-shared-2.4.19-386-udeb socket-modules-2.4.19-386-udeb
isa-pnp-modules-2.4.19-386-udeb nic-modules-2.4.19-386-udeb
Total system size: 2.5M (1.4M libs, 356K kernel modules)
Initrd size: 970k
Kernel size: 656k
Free space: -186k



msg25813/pgp0.pgp
Description: PGP signature