Comment 16 for bug 719129

Revision history for this message
Curtis Gedak (gedakc) wrote :

Since we have several topics of discussion, I will try to break these
into sections. This will hopefully make easier to see what we need to
do. :-)

    ------------------------------
1) THANK YOU FOR YOUR PATCHES AND INTEREST IN PARTED AND GPARTED

I wish to begin by thanking you for the patches you have supplied for
parted. Your patches along with another from Jim Meyering have been
critical in helping to lessen problems for users of Parted and
GParted arising from a "failure to inform kernel of partition
changes".

The Jim Meyering patch I speak of is this one:
http://git.debian.org/?p=parted/parted.git;a=commit;h=0f850220b3f26bb969a1a7ff78dc550691a89566

There may yet be some further room for improvement in this area
because I have at least one report of the problem continuing.
See:
http://gparted-forum.surf4.info/viewtopic.php?id=14357
Personally I have not been able to reproduce this problem.

    ------------------------------
2) WHAT TO DO FOR FOR THE UPCOMING UBUNTU NATTY RELEASE?

Details of the changes in each release of GParted can be found in the
NEWS file:
http://git.gnome.org/browse/gparted/tree/NEWS

If it is too late in Natty's release cycle to include a newer version
of GParted, then perhaps we should focus on NATTY+1. The possibility
of duplicate dmraid device partition names has been present since
DMRaid support was added in GParted 0.4.4.

If there is still time to consider patches in Natty, I suggest the
following patch to fix incorrect dmraid partition path names:
http://git.gnome.org/browse/gparted/commit/?id=3c35a7ff42927bfcd89465661766b661f4fbcb76
In my mind this problem has a higher possibility of impacting a user
than duplicate dmraid device names.

    ------------------------------
3) WHY MAINTAIN BACKWARD COMPATIBILITY WITH OLDER LIBPARTED VERSIONS?

You might call this a personal pet peeve, but I hate when I am forced to
upgrade my entire operating system just to run a newer version of an
application. I find this particularly vexing when the operating
system or GNU/Linux distribution is still within a supported
time-frame.

This is why I _try_ my best to ensure that GParted will compile and
run on the currently supported major GNU/Linux distributions. I
recognize that there are situations where it makes sense to break from
this goal.

Unfortunately the following statement is not always true:
"If the distribution can update to the latest version of gparted, then
they can update parted as well just as easily."

A case in point is with Ubuntu 8.04 Hardy Heron. This release is
currently supported on both the desktop and server. However,
libparted-2.3 does not work well with GParted on Ubuntu 8.04. The
problem encountered when performing operations in GParted with
libparted 2.3 is the unfortunately familiar "failure to inform kernel
of partition changes".

Older versions of libparted, such as 1.7.1 that is packaged for Ubuntu
8.04, do not exhibit this problem on Ubuntu 8.04.

I have tested this on a physical computer installation, and in a
virtual machine with the same results.

    ------------------------------
4) WHAT TO DO FOR THE UPCOMING UBUNTU NATTY+1 RELEASE?

I would like to work together with you to accomplish the following
three goals:

    a) Begin the retirement of DMRaid.cc
    b) Resolve the duplicate dmraid device partition name issue
    c) Use libparted-2.3 for dmraid device partition naming

For these goals, perhaps you could help me to help you. I know
exactly where to place the configure conditionals. However, I am
searching for a more elegant way to set a configure condition such as
"HAVE_LIBPARTED_2_3_0_PLUS".

Currently I have duplicated code (I know this is not desirable), to
determine if libparted is >= 1.7.1, and also to determine if libparted
is >= 2.2. See:
http://git.gnome.org/browse/gparted/tree/configure.in#n43

If you can help reduce this duplication of code then I would be very
appreciative.

As I stated above, I know exactly where to place the #ifdef
HAVE_LIBPARTED_2_3_0_PLUS conditionals to accomplish the above goals.

Hopefully that helps to address questions raised, and outline a path forward.

Sincerely,
Curtis Gedak