[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Some RC bug work

Frank Lichtenheld <djpig@debian.org> writes:

> You're doing great AFAICT.

Excellent.  Then I will keep going.  :)  I have another message I'm
getting ready to send out (I'm up to x on the list).

>> #256093  interchange: bug in package logic requiring prerm workaround

> this sounds klunky, leaving for now

Just to fill in a few more details....

The package bug fix is very straightforward.  If the server isn't running,
stop and don't try to do anything further with killing it.  The current
package will continue on with a server PID of 0 and then proceed to kill
its own interface script with a kill -HUP 0, thus confusing the postinst
and prerm scripts.

The klunky bits are working around the fact that this bug is present in
woody and will be triggered by the woody prerm script on an upgrade to
sarge.  I therefore needed to do two things:

 * Add an || /bin/true to the end of the invocation of interchange in the
   prerm script in the sarge package, so that when the woody prerm fails
   and dpkg falls back to running the new prerm script, it will succeed.
   This solves the initial problem.

 * Add a versioned dependency of interchange-cat-foundation on whatever
   version contains the fix to interchange explained above.  I didn't do
   this in a patch since I wasn't sure what version that would be, but
   it's easy to add to control.  This is needed to also ensure that,
   during an upgrade, interchange-cat-foundation's postinst won't run
   before interchange has been upgraded (since otherwise it will have the
   same problem).

So yeah, agreed, it is a bit complicated to analyze.  (I'll add the above
to the bug as well, since I think it's a better explanation than I put in
there the first time.)

Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Reply to: