Re: strange preinst tests about --assert-working-epoch
On Fri, Feb 15, 2002 at 12:22:33AM +0100, Raphael Hertzog wrote:
> Le Thu, Feb 14, 2002 at 02:07:47PM -0800, Martin Quinson écrivait:
> > -- Michael Alan Dorman <email@example.com> Sun, 26 Jan 1997 16:49:19 -0500
> > and he's right, maintaining compatibility with old boxes is important.
> Did you see the date ? 1997 ! I suggest you don't loose time to write
> a lintian test for checking a bug that used to be frequent 5 years
> ago ...
Yes I did. And I put it in my mail on purpose.
> *Nobody* is going to use a dpkg that doesn't handle epochs. Epochs have
> been implemented since may 1996 ... in 1997 it could happen to use a
> dpkg of more than a year old, but now 4 releases later and 5 years
> later, nobody is using a dpkg << 1.2.0.
I completely agree. I didn't meant that all the packages with epoch should
check for it. I more wanted to get rid of the cruft in the packages doing
the test. Why ? Well, for no real reason, in fact.
I recently discovered that it is possible to use remote databases in
debconf, making it a really great tool to managing clusters of machines. But
it would work even better if all packages use this system. So, I begun
grepping around in /var/lib/dpkg/info for a package still using direct
interaction with user during install, and small enough so that I can trust
myself to propose a debconfized patch to the maintainer. On the way, I did
found this strange and crufty (in my point of view) check, and reported it.
Nothing critical, in fact ;)
Si les grands esprits se rencontrent, les petits esprits, eux, se cognent.