Re: testing to be implemented on ftp-master

[Brian May <bam@debian.org>]

>>>>> "Anthony" == Anthony Towns <aj@azure.humbug.org.au> writes:

    Anthony> That's the other problem. At the moment, it's just
    Anthony> assumed that all open RC bugs apply to the latest version
    Anthony> in unstable, and that when they no longer apply they get
    Anthony> closed.

So, am I correct in assuming that if a RC bug exists for a package,
then it will never get into testing?

Hmmm... I dislike the current BTS design of only keeping track of the
initial version of the package that creates the bug. True - the system
of having tags to include the distribution helps a little bit, but,
IMHO, it would be much cleaner if the BTS simply recorded some sort of
open ended list of what versions of the package the bug applies
to. Plus, as an extra bonus, it would always work, no matter how many
stages exist between unstable and stable. It would also work if I
wanted to create and maintain my own custom distribution, eg from
stable and unstable packages.


I will also comment on a non-so obvious issue. Hopefully this will
never occur, but just in case somebody hasn't thought about it,
consider:

1. existing package version 1 in unstable has a serious security
bug that will result in all files being erased on the computer.

2. maintainer uploads a new version 2 with the bug fixed.

3. dinstall (assuming it is still called dinstall) closes the bug and
tries to install the package in unstable. Unfortunately, for some
reason, it fails to install the package (eg disk error, package error,
etc). Of course, if closing bugs is not done until the last step, this
problem wont occur.

4. The relevant program sees version 1 in unstable meets the required
criteria for testing, and installs it in testing.

-- 
Brian May <bam@debian.org>