Bugzilla Bug 143485

There was an API change in libreiserfs from progsreiserfs-3.0 to
progsreiserfs-3.1_rc8.  testdisk needs the library from 3.1 (according to its
ebuild) however sys-apps/parted (and hence qtparted, which is where I came in,
and presumably everything else that uses parted) needs the library from 3.0.4.

Note that libreiserfs is a "soft" dependency - i.e. instead of being listed in
libparted's as "NEEDED", it is referenced via libdl when required.  So it's not
until you run a utility that uses libparted and then tries to do something with
a reiserfs partition that you get the error from libparted written to stderr.

Can we re-instate 3.0.4 (as a library-only ebuild, obviously) and slot the two
library versions?

The API version is defined by LIBREISERFS_INTERFACE_AGE (its "minimum" version)
and LIBREISERFS_BINARY_AGE (its "maximum" version).  These are both 0 on 3.0.4,
and both 1 on 3.1_rc8.

I don't know if that will be enough; it may be that libparted isn't smart
enough to find the right libreiserfs version, but one step at a time.

It's not maintained upstream since ages and known to trash reiserfs partitions.
See bug 84514.

According to that bug it's the apps that were broken, rather than the library,
and the apps have long since been discarded (parted just uses the library).

this package is complete garbage and needs to die

if someone else wants to take it, feel free

punting from base-system

vapier: The last time progsreiserfs came up, I pointed all of this out, and
noted that I was taking care of it, under my work in base-system. I'll add
myself to the maintainer list, but I want the bugs to go to base-system still,
so I check them with other core stuff.

kevquinn: I strongly advise against using progsreiserfs with parted. In
specific, the parts that are broken are create/resize/cpfs. testdisk uses
progsreiserfs for data recovery only, so it's safe there.

parted SHOULD migrate to using the commandline reiserfsprogs instead of
progsreiserfs, but you'll have to get upstream to do that.

i'm tired of seeing this broken ass package, i would just as soon punt it

The single reason that I want progsreiserfs kept around is that combined with
testdisk it provides the ONLY reliable way to recover data from damaged
reiserfs partitions.

I'm perfectly ok with you banning any other use of it, if you can figure out
how to do so.

(In reply to comment #4)
> kevquinn: I strongly advise against using progsreiserfs with parted. In
> specific, the parts that are broken are create/resize/cpfs. testdisk uses
> progsreiserfs for data recovery only, so it's safe there.

Adding allanonjl as maintainer of sys-apps/parted and livecd herd to CC.  If
sys-apps/parted has the use of write parts of libreiserfs stripped out, my
stuff (qtparted) will pick that up automatically.

I searched around for information about problems with progsreiserfs, but the
author (as recently as mid 2005) asserted he's had no bug reports of any
substance against progsreiserfs, and that as far as he was concerned the
problems were down to the apps that used it rather than the library.  Of course
others disagree, but it seems that no-one has actually posted detailed fault
reports.  If sys-apps/parted is modified to eliminate use of progsreiserfs, it
would be useful to find and record such information.

For the record; here is some discussion on the GNU parted mailing list:

http://lists.gnu.org/archive/html/bug-parted/2004-03/msg00016.html
http://lists.gnu.org/archive/html/bug-parted/2002-12/msg00016.html
http://lists.gnu.org/archive/html/bug-parted/2004-10/msg00012.html

The situation is not clear.  Some historical (2004) problems that Debian
suffered from, according to progsreiserfs maintainer were down to patches
Debian applied (which broke big-endian machines).

One recommendation that comes up frequently is to do run reiserfsck (from
reiserfsprogs) with options like -rebuild-sb, -rebuild-tree to ensure the
filesystem is consistent first.  In v0.3.0.6 Yuri (progsreiserfs author) added
some tree checks to the resize stuff.  Yuri has stated several times that if
the filesystem is already broken (i.e. inconsistent tree), running tools that
use libreiserfs over it (e.g. resize etc) will likely make it worse.

And it's still breaking stuff; punt this junk finally.

Jakub: Hello? see that I say parted people should switch to using the
commandline reiserfsprogs tools, and NOT use any of the modification stuff in
progsreiserfs, and that progsreiserfs needs to stay for it's data-recovery
stuff used by testdisk.

(In reply to comment #10)
> Jakub: Hello? see that I say parted people should switch to using the
> commandline reiserfsprogs tools, and NOT use any of the modification stuff in
> progsreiserfs, and that progsreiserfs needs to stay for it's data-recovery
> stuff used by testdisk.
> 

Yeah, hello indeed... The progsreiserfs junk breaks qtparted/gparted/parted
whenever *installed* (Bug 84514), not just when something's compiled against it
- yeah you can easily test this by installing parted, then installing
progsreiserfs and then trying to run parted... 

So -  I really don't see how people should switch. It plain needs to die, it's
dead, unmaintained, broken and makes baby jesus cry. If something needs to be
fixed, then it's testdisk that shouldn't use this.

who changed parted back to using it?
2 years ago we changed parted AWAY from using it.

the root of it is that progsreiserfs has TWO sets of functionality.
1. data-recovery of reiserfs partitions
2. creation/maintaince of reiserfs partitions

testdisk ONLY uses #1, and there is absolutely nothing else out there that
provides that ability.
parted ONLY uses #2, but the ability is also available from the official
reiserfsprogs.

(In reply to comment #12)
> who changed parted back to using it?
> 2 years ago we changed parted AWAY from using it.

Erm, still misunderstanding each other? I'm not suggesting that parted should
be changed back to using sys-fs/progsreiserfs; I'm suggesting that
sys-fs/progsreiserfs should finally *bite the dust* and nothing should use it
at all... At minimum it should be dropped from qtparted and should block
*parted because it plain breaks it.

You said this:
"The progsreiserfs junk breaks qtparted/gparted/parted
whenever *installed*"
This bug was fixed years ago already, I'm wondering why it came back.

Removing progsreiserfs leaves no way for those poor users of reiserfs to
recover their data with testdisk so they can migrate away from it ;-)

(In reply to comment #14)
> Removing progsreiserfs leaves no way for those poor users of reiserfs to
> recover their data with testdisk so they can migrate away from it ;-)

They can always pay $25 to Hans' legal expenses fund and get their answer? :P 

Mkay, I guess I won't make you punt this, so how about the blocker? :)

parted needes to drop reiserfs usage entirely for the moment, since it's
upstream isn't doing anything on moving to pure mkfs (reiserfsprogs), and
progsreiserfs should be used less in future.

debian has entirely dropped reiserfs support from testdisk, and while we won't
go that far, nothing other than testdisk should be using progsreiserfs.