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

Bug#376590: developers-reference: suggests to test downgrading

On 03/07/06 at 16:18 -0400, Justin Pryzby wrote:
> Package: developers-reference
> Version: 3.3.7
> Severity: normal
> |* Downgrade the package to the previous version (if one exists) ---
> |this tests the `postrm' and `prerm' scripts.
> Although downgrading *might* be actively supported by some packages,
> this seems to imply that it must be supported, which encourages people
> to think that they're allowed to downgrade stuff.

Agreed, I think that this line could be removed.

> Also, it seems misleading in its suggestion about the
> {post,pre}removal maintscripts, which are also exercised during any
> upgrade (though different paths are not, and it takes pretty
> deliberate effort to test eg. the abort-* and failed-* paths).
> A better recipe for testing (copied from a message to my AM):
>   . put package into "old version installed" state, perhaps by purging
>     whatever's installed, and apt-get installing some version from the
>     archive (probably shouldn't be older than the most recent stable
>     release).
>   . dpkg -i foo.deb
>   . check that the upgrade happend as intended; this might include
>     things like: conffile updates, /usr/doc transition stuff,
>     symlink/dir conversion stuff, etc.  Basically ensuring that the
>     maintscripts do what's intended.  The package should of course be
>     usable :)
>   . dpkg -r foo
>   . again make sure the maintscripts do what's intended.  Make sure
>     that eg. info documents and alternatives are unregistered.
>   . dpkg -P foo
>   . make sure that any generated files (eg. in /var/{lib,cache} and
>     /etc/) are removed.  This also includes debconf cache and anything
>     similar eg. update-rc.d removal.
>   . dpkg -i foo.deb
>   . Check more maintscript actions; the package should be usuable
>   . dpkg -i foo.deb
>   . The package should still be usable :)  This should probably just
>     check for idempotency at some level; eg. that installation doesn't
>     depend on some file that is only created in preinst.

It sounds too long to include in devref as-is. Could you maybe shorten
it a bit and provide a patch against section 5.3?
| Lucas Nussbaum
| lucas@lucas-nussbaum.net   http://www.lucas-nussbaum.net/ |
| jabber: lucas@nussbaum.fr             GPG: 1024D/023B3F4F |

Reply to: