David Kalnischkies <kalnischkies+deb...@gmail.com> writes:
> 2010/4/20 Goswin von Brederlow <goswin-...@web.de>:
>> David Kalnischkies <kalnischkies+deb...@gmail.com> writes:
>>> 2010/4/18 Goswin von Brederlow <goswin-...@web.de>:
>>>> And even if the script knows about it the setting can not be overriden.
>>> The config (sub)trees can be removed with the clear command in
>>> a configuration file, so e.g.
>>> #clear APT::Update::Pre-Invoke;
>>> will get right of all specified options in this tree.
>>
>> #clear APT; clears the APT subtree.
>> #clear ; does nothing.
>>
>> Shouldn't the later clear the whole tree?
> A clear-all isn't implemented - and given that it would have been
> the same result as the APT_CONFIG trick we don't need it - do we?
Need no. But then it could give an error.
>>>> --no-default-config
>>>> Do not read the default configuration file from /etc/apt/.
>>> ( I assume you mean "files" here and therefore not only the "main"
>>> apt.conf but also all files in apt.conf.d. )
>>> What can be achieved also by setting a config file with APT_CONFIG
>>> environment variable and in that file the paths for the configuration files.
>>
>> That might be worth adding to "man apt-get". If it indeed works to
>> replace the default conffile then that is enough.
> It is described nowhere, but can be "extracted" from the apt.conf
> manpage as in the DESCRIPTION section there is described in which
> order the configuration files and options are parsed/evaluated.
>
>
>> E.g. when you build a debian-installer for i386 you do not to
>> suddenly get amd64 packages mixed in because the user enabled multiarch
>> in the system wide config.
> These systems already have a problem with preferences and sources.list -
> or do they accept these as valid configuration while the complete config
> is invalid? They are better of using their complete own apt setup (Dir::)
> than fighting with disabling specific option, yes. I thought more of smaller
> scripts/applications operating on the "normal" system which just can't
> cope with $random option (yet).
That is the probelm I run into. Currently all the implementations I've
seen use either -c or -o to override known options, like use their own
sources.list file. Every option they forget to override is a potential
problem. None of them used '#clear ...;'.
I too believe the right thing for them is to use APT_CONFIG to completly
replace the config and start defining things from scratch like you say.
I will have to write a patch for D-I and ia32-libs*.
>> I agree though that the bug can be closed if APT_CONFIG is documented
>> more publicly. E.g. add a reference to it in the -c option. If there is
>> already a way to not use the default conffile then we don't need another
>> way.
> How about
>>>>>>
> -c, --config-file
> Configuration File; Specify a configuration file to use.
> The program will read the default configuration file and then this
> configuration file. If configuration settings need to be set before the
> default configuration files are parsed specify a file with the APT_CONFIG
> environment variable. See apt.conf(5) for syntax information.
> <<<<<
> ?
Perfect. Add that and I consider the bug closed.
MfG
Goswin
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
