Hi Wouter and Jean Baptiste,
On Wed, Nov 19, 2014 at 11:35:50PM +0100, Wouter Verhelst wrote:
> On Tue, Nov 18, 2014 at 01:48:04PM +0100, Jean Baptiste Favre wrote:
> > * Remove debconf calls from ola-rdm-tests postinst. (Closes: #767676)
> > * Fix other seriouys issues:
> > - Provides missing /etc/default/ola from ola postinst script to allow
> > olad service control in the same way rdm_test_server is
>
> NAK. /etc/default/* to manage service status is a menace, and should be
> forbidden.
Wouter, please bear in mind, that it is not the proposed NMU that is
adding the need for an /etc/default/ola file. The current init script
(debian/ola.olad.init) sources /etc/default/ola and doesn't start at all
when it does not exist. This puts you in a bad position to demand
otherwise.
Given the other things you said, I assume that it is not your intention
for the init script to check for $RUN_DAEMON at all. (The issue is
duplicated in the init script of ola-rdm-tests.)
I suggest the following approach then:
* Remove the check for $RUN_DAEMON from both init scripts (olad and
rdm_test_server).
* Do not ship the default files in either package.
* Keep sourcing /etc/default/ola{,-rdm-tests} so users can change e.g.
$DAEMON_ARGS.
Optionally:
* During upgrade, check whether the default files were unmodified (by
comparing their md5sum to known values) and if they are unmodified,
remove them.
Wouter, do you agree with this approach?
Jean Baptiste, would you like to invest further work into this?
The "unowned file after purge issue" has bug number #769331 and should
be mentioned in the changelog.
Since I am building in pbuilder, #769670 will need fixing as well. What
happens there is that pbuilder creates a new network namespace (using
"unshare -n" from util-linux) and runs the build with just lo being
available. The source code for the test is as follows.
common/network/InterfacePickerTest.cpp:
| auto_ptr<InterfacePicker> picker(InterfacePicker::NewPicker());
| vector<Interface> interfaces = picker->GetInterfaces(false);
| OLA_ASSERT_TRUE(interfaces.size() > 0);
The false value in the second line is bound to a parameter called
"include_loopback". So the test examines all interfaces, ignores "lo"
and errors out, because it didn't find any (other than lo). The test
simply doesn't make any sense. It verifies a property that does not hold
in general.
Given the zoo of issues in ola, I am not convinced that it is ready for
being part of a stable release. It didn't build on any buildd either
(see #755866). I am niether convinced that the "missing source" issue
(#760414) really is non-RC, because the source for the minified
jquery-ui library is not in the Debian archive. It is only available in
a different version and we have previously declared "source is only
available in a different version" as RC buggy.
Helmut
--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
