Bacula Admin Tool (bat) does not work well with Qt 4.6.2 in Lucid

[Etienne Goyer]

Binary package hint: bacula-console-qt

On lucid, in the Version Browser of bat, it is impossible to check the latest version of a file. This is described in the following upstream bug report: http://bugs.bacula.org/view.php?id=1550 (login using anonymous/anonymous)

For convenience, here is the verbatim of the bug description:

--------------
I often restore a single file, so I use Version Browser.

1) I browse for the file and when clicking on it, I see many different versions of it.
2) When I try to mark the file (tick the checkbox), it cannot be done
3) When I try to mark the most recent version of the file in 'Versions of File', it cannot be done
4) When I try to mark any other version of the file, it works OK
5) When there's another version of the file available (after the next backup), the version which couldn't be marked, can now be marked and the most recent can not

It seems to be so for every job and every file within a job.

I turned on all the debugging options of Bat, but there's nothing said when marking (or trying to) files/versions.
-----------------

Response from the lead developer of Bacula:

-----------------
I am unable to reproduce this. It works fine with my version of bat. You are using a bat that was most likely compiled with Qt 4.6.2, and bat is known not to work with all versions of Qt because they are, unfortunately, not compatible.

You must use a bat compiled with Qt 4.3.4. All the project binaries are compiled with Qt 4.3.4 using the released depkgs-qt code.
-----------------

Lucid indeed ship with Qt 4.6.2. I can confirm that bat does not show this problem on maverick, which ship Qt 4.7.0. To confirm the problem is related to the version of Qt being used, I backported bacula-console-qt (5.0.2) from maverick to lucid. The resulting binary showed the same bug as the bacula-console-qt package from lucid.

[Kern Sibbald (kern)]

The new version of Bacula (5.2.0) to be released in about a month has a number
of fixes, and will work perfectly with Qt 4.6.2 (Lucid) where I have done the
development work. There have been quite a number of enhancements and a lot
of stability improvements in bat 5.2.0.

As I previously reported, bat uses Qt, which has some important compatiblity
problems between versions. Hopefully the new 5.2.0 bat will also work on
Maveric. If not, I will be releasing a depkgs-qt-4.6.2 that can be used to
compile bat on any Ubuntu release.

Regards,

Kern

[Clint Byrum (clint-fewbar)]

Kern, does this mean that there is an explicit fix in the code of 5.2.0 that will make this problem go away on Qt 4.6.2 ?

While I do think we should put 5.2.0 in backports when it arrives, if the fix is simple it might be worth patching 5.0.1 so it works in lucid.

This one is really hard to confirm as I don't have any files to restore in my test bacula. Is there a shorter test case?

I'm going to mark this as Incomplete, pending a shorter test case and/or answers from Kern. If its too hard to test, and/or too hard to fix in lucid, it might just have to be a "Won't Fix".

[Clint Byrum (clint-fewbar)]

Actually since it has been confirmed upstream I'm marking it as Confirmed, Importance Medium, since there doesn't seem to be a workaround.

[Kern Sibbald (kern)]

On Saturday 02 April 2011 02:23:21 Clint Byrum wrote:
> Kern, does this mean that there is an explicit fix in the code of 5.2.0
> that will make this problem go away on Qt 4.6.2 ?

There have been a very large number of changes to bat from 5.0.x to 5.2.0,
including new features, bug fixes, and particularly seg fault fixes. I am
sorry, I don't remember the specifics, and since I did not explicitly
reproduce the reported bug, I am not sure there is an explicit fix, but I
think so.

>
> While I do think we should put 5.2.0 in backports when it arrives, if
> the fix is simple it might be worth patching 5.0.1 so it works in lucid.

There were a lot of changes, and I believe we have a significant number in the
Director too, so any backporting might not be easy. You are free to try
backporting but it is not something that I am personally going to do.

>
> This one is really hard to confirm as I don't have any files to restore
> in my test bacula. Is there a shorter test case?

I don't know as I haven't explicitly reproduced the reported bug.

>
> I'm going to mark this as Incomplete, pending a shorter test case and/or
> answers from Kern. If its too hard to test, and/or too hard to fix in
> lucid, it might just have to be a "Won't Fix".

OK. I am not sure what your procedure is -- it might be good to note "fixed
in 5.2.x" providing that is really the case.
* Changed in: bacula (Ubuntu)
> Status: Incomplete => Confirmed

[Clint Byrum (clint-fewbar)]

Excerpts from Kern Sibbald's message of Sat Apr 02 07:20:23 UTC 2011:
> On Saturday 02 April 2011 02:23:21 Clint Byrum wrote:
> > Kern, does this mean that there is an explicit fix in the code of 5.2.0
> > that will make this problem go away on Qt 4.6.2 ?
>
> There have been a very large number of changes to bat from 5.0.x to 5.2.0,
> including new features, bug fixes, and particularly seg fault fixes. I am
> sorry, I don't remember the specifics, and since I did not explicitly
> reproduce the reported bug, I am not sure there is an explicit fix, but I
> think so.
>
> >
> > While I do think we should put 5.2.0 in backports when it arrives, if
> > the fix is simple it might be worth patching 5.0.1 so it works in lucid.
>
> There were a lot of changes, and I believe we have a significant number in the
> Director too, so any backporting might not be easy. You are free to try
> backporting but it is not something that I am personally going to do.
>
> >
> > This one is really hard to confirm as I don't have any files to restore
> > in my test bacula. Is there a shorter test case?
>
> I don't know as I haven't explicitly reproduced the reported bug.
>
> >
> > I'm going to mark this as Incomplete, pending a shorter test case and/or
> > answers from Kern. If its too hard to test, and/or too hard to fix in
> > lucid, it might just have to be a "Won't Fix".
>
> OK. I am not sure what your procedure is -- it might be good to note "fixed
> in 5.2.x" providing that is really the case.

Hm. Sounds like this one may not get backported.

The procedure is to mark the main bug task Fix Released, but then track
each release of the package before that individually. Won't Fix means the
bug is in that version, but we're not going to fix it.

> * Changed in: bacula (Ubuntu)
> > Status: Incomplete => Confirmed
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/735369
>
> Title:
> Bacula Admin Tool (bat) does not work well with Qt 4.6.2 in Lucid
>
> Status in Bacula the Network Backup Solution:
> New
> Status in “bacula” package in Ubuntu:
> Confirmed
>
> Bug description:
> Binary package hint: bacula-console-qt
>
> On lucid, in the Version Browser of bat, it is impossible to check the
> latest version of a file. This is described in the following upstream
> bug report: http://bugs.bacula.org/view.php?id=1550 (login using
> anonymous/anonymous)
>
> For convenience, here is the verbatim of the bug description:
>
> --------------
> I often restore a single file, so I use Version Browser.
>
> 1) I browse for the file and when clicking on it, I see many different versions of it.
> 2) When I try to mark the file (tick the checkbox), it cannot be done
> 3) When I try to mark the most recent version of the file in 'Versions of File', it cannot be done
> 4) When I try to mark any other version of the file, it works OK
> 5) When there's another version of the file available (after the next backup), the version which couldn't be marked, can now be marked and the most recent can not
>
> It seems to be so for every job and every file...

[Kern Sibbald (kern)]

On Saturday 02 April 2011 16:40:13 Clint Byrum wrote:
> Excerpts from Kern Sibbald's message of Sat Apr 02 07:20:23 UTC 2011:
> > On Saturday 02 April 2011 02:23:21 Clint Byrum wrote:
> > > Kern, does this mean that there is an explicit fix in the code of 5.2.0
> > > that will make this problem go away on Qt 4.6.2 ?
> >
> > There have been a very large number of changes to bat from 5.0.x to
> > 5.2.0, including new features, bug fixes, and particularly seg fault
> > fixes. I am sorry, I don't remember the specifics, and since I did not
> > explicitly reproduce the reported bug, I am not sure there is an explicit
> > fix, but I think so.
> >
> > > While I do think we should put 5.2.0 in backports when it arrives, if
> > > the fix is simple it might be worth patching 5.0.1 so it works in
> > > lucid.
> >
> > There were a lot of changes, and I believe we have a significant number
> > in the Director too, so any backporting might not be easy. You are free
> > to try backporting but it is not something that I am personally going to
> > do.
> >
> > > This one is really hard to confirm as I don't have any files to restore
> > > in my test bacula. Is there a shorter test case?
> >
> > I don't know as I haven't explicitly reproduced the reported bug.
> >
> > > I'm going to mark this as Incomplete, pending a shorter test case
> > > and/or answers from Kern. If its too hard to test, and/or too hard to
> > > fix in lucid, it might just have to be a "Won't Fix".
> >
> > OK. I am not sure what your procedure is -- it might be good to note
> > "fixed in 5.2.x" providing that is really the case.
>
> Hm. Sounds like this one may not get backported.
>
> The procedure is to mark the main bug task Fix Released, but then track
> each release of the package before that individually. Won't Fix means the
> bug is in that version, but we're not going to fix it.

OK, your procedures are different from ours, which is OK.

We normally indicate to the user that it is fixed in a future version. That
is all I was trying to say. No problem however you do it :-)

>
> > * Changed in: bacula (Ubuntu)
> >
> > > Status: Incomplete => Confirmed
> >
> > --
> > You received this bug notification because you are a direct subscriber
> > of the bug.
> > https://bugs.launchpad.net/bugs/735369
> >
> > Title:
> > Bacula Admin Tool (bat) does not work well with Qt 4.6.2 in Lucid
> >
> > Status in Bacula the Network Backup Solution:
> > New
> > Status in “bacula” package in Ubuntu:
> > Confirmed
> >
> > Bug description:
> > Binary package hint: bacula-console-qt
> >
> > On lucid, in the Version Browser of bat, it is impossible to check the
> > latest version of a file. This is described in the following upstream
> > bug report: http://bugs.bacula.org/view.php?id=1550 (login using
> > anonymous/anonymous)
> >
> > For convenience, here is the verbatim of the bug description:
> >
> > --------------
> > I often restore a single file, so I use Version Browser.
> >
> > 1) I browse for the file and when clicking on it, I see many different
> > versions of it. 2) When I try to mark the file (tick the checkbox), it
> > cannot be done 3) When I try to mark th...

[Clint Byrum (clint-fewbar)]

Excerpts from Kern Sibbald's message of Sat Apr 02 15:47:37 UTC 2011:
> On Saturday 02 April 2011 16:40:13 Clint Byrum wrote:
> > Excerpts from Kern Sibbald's message of Sat Apr 02 07:20:23 UTC 2011:
> >
> > Hm. Sounds like this one may not get backported.
> >
> > The procedure is to mark the main bug task Fix Released, but then track
> > each release of the package before that individually. Won't Fix means the
> > bug is in that version, but we're not going to fix it.
>
> OK, your procedures are different from ours, which is OK.
>
> We normally indicate to the user that it is fixed in a future version. That
> is all I was trying to say. No problem however you do it :-)
>

Right, we don't really have the luxury of putting people on new upstream
releases, though we'll indicate the status of the bug upstream so users
can go and get the new version if they want.

It *would* be good to link this to lucid-backports once 5.2.0 releases
so people can at least see that this bug would be fixed by a backport
of bacula 5.2.0.