Bug#858187: boinc-manager: boincmgr suddenly needs -d /etc/boinc-client

[Christian Beer]

On 04.04.2017 09:00, Russell Coker wrote:
> $ boincmgr -n localhost
> [fails]
> $ boincmgr -n localhost -d /etc/boinc-client
> [works]
> $ dpkg -l boinc-manager|cat
> Desired=Unknown/Install/Remove/Purge/Hold
> | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
> |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
> ||/ Name           Version        Architecture Description
> +++-==============-==============-============-
> ================================================
> ii  boinc-manager  7.6.33+dfsg-10 amd64        GUI to control and monitor the 
> BOINC core client
>
I think I know what's going on. If you start boincmgr there are two
places in the code that try to determine the location of the data and
configuration directory. We fiddled with one of those places to fix the
multiple Manager issue fixing this case. I think the previous behavior
was not intended but worked somehow.

The problem is that starting boincmgr from an arbitrary directory makes
it hard for the Manager to determine where the configuration files are
(it only looks in the current working directory). The assumption is that
boincmgr is always started in the directory where the configuration
files are (aka data directory). Using the menu links this is
automatically taken care of by changing the working directory before
starting boincmgr.

What I could imagine is a Debian specific fallback to the data directory
detection logic that first tries the current directory and if no config
files are found tries /var/lib/boinc-client/ which would restore the
previous behavior to boincmgr -n localhost that makes it independent
from the where it is started from. What I would like to investigate too
is the distinction between the -d and -e flags and what internal
variables they control in order to not create a regression and brake
something else.

Regards
Christian