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

Bug#517644: [amd64][lenny] since lenny release, unable to install 'etch' via 'lenny' installer



reassign 517644 choose-mirror
severity 517644 serious
thanks

On Sunday 01 March 2009, Vincent McIntyre wrote:
> Comments/Problems:
> I attempted to install 'etch' on this system using the 'lenny'
> installer. I did this via the preseeding method, passing the
>   suite="etch"
> option to the boot line of the installer kernel.
> The system then proceeds to install 'lenny'.
> This appears to be a regression relative to the situation before
> release, when I was able to successfully "cross-install" etch via the
> lenny installer.

As this is the second report of this issue I've seen passing by, it's 
quite likely this is valid.

This is definitely a serious issue that should be fixed with the next 
stable update. It should also be verified whether oldstable installs 
using Sarge netboot images still work or not. There's a good chance those 
are broken too.

> In the preseed I have set the "etch" distribution as well
> % grep -w etch /var/lib/preseed/log
> d-i mirror/suite select etch
> d-i preseed/url string
> http://installserver/./preseed/debian/etch/amd64/server-lenny.cfg
> d-i mirror/suite            string etch
> d-i mirror/codename         string etch

There's some weirdness in here (double definition of mirror/suite for 
example), but I doubt that affects the validity of the report.

> I happen to use an apt-proxy but I don't think that is the cause -
> doing the first wget manually returns the expected value of Suite:,
> i.e. oldstable, and if I do the next one manually, with 'oldstable' in
> the URL I get the right Codename returned ('etch').

Agreed, but it should be confirmed with a real mirror just to make sure.

> I also noticed that 'packages/choose-mirror/mirrors.h' does not include
> "oldstable" in the definition of the suites[][SUITE_LENGTH] array.
> Could this be the issue?

It could well be. This is something we have been unable to test so far. I 
had expected it to work as we do use the codename as a base to determine 
the suite, but if there is a check that the returned suite must be in 
that list the failure would be explained.



Reply to: