svn commit: r275219 - projects/building-blocks/share/skel

[Adrian Chadd]

.. is vi marked as a tool required for building (ie, ex) ?


-a


On 30 November 2014 at 20:27, Garrett Cooper <yaneurabeya at gmail.com> wrote:
> On Nov 30, 2014, at 12:25, Jilles Tjoelker <jilles at stack.nl> wrote:
>
>> On Fri, Nov 28, 2014 at 10:07:16PM +0000, Garrett Cooper wrote:
>>> Author: ngie
>>> Date: Fri Nov 28 22:07:16 2014
>>> New Revision: 275219
>>> URL: https://svnweb.freebsd.org/changeset/base/275219
>>
>>> Log:
>>>  Install skeleton files based on the values of MK_MAIL and MK_TCSH
>>
>>>  Customize according to MK_GAMES and MK_VI, in particular comment out lines
>>>  that would change behavior
>>
>>>  Change dot.cshrc and dot.profile to not check for /usr/games/fortune based
>>>  on the new behavior in customizing the files
>>
>> Hmm, this assumes a rather specific use case: selecting a particular
>> combination of options, building everything and installing everything
>> what was built.
>>
>> In particular, the run-time check for /usr/games/fortune was there
>> because a user might not install /usr/games, even though it was part of
>> the built release. With "packaged base", this possibility will remain
>> important.
>
> Good point.
>
> What triggered this commit was something I noticed when I fixed “MK_VI == no” to delete vi in base, which in turn breaks the default dot files, i.e. 'export EDITOR=vi’. MK_VI == no doesn’t install vi with installworld, but it doesn’t clean it up on head if you run make delete-old either, so this issue wasn’t apparent until I started this work (and I guess no one complained about the MK_VI == no case before now). To err on the safe side, I can add a `if [ -x /usr/bin/vi ]; then ; fi` around the export EDITOR line if you like — however many other OS distros (Fedora Linux, IllumOS, Ubuntu Linux, etc) don’t ship with defaults — is this something we should do as well?
>
> vi is also implicitly required for the build when building some files in share/termcap, as the build calls ex on the files, but that’s a sidenote.
>
>> Also, dotfiles might be shared between systems by users.
>
> Yes, but shouldn’t the person customizing the system know what files are being distributed?
>
>> This is an objection against customizing dot.* file contents depending
>> on options, not against conditionally installing dot.* files.
>
> Ok. Something else that needs to be discussed is /usr/games being in $PATH and a few other things, because this could slow down path lookups, even if it’s a small amount (I realize the namei cache is supposed to make this less noticeable, but every little bit counts IMHO).
>
> Thanks for the input :)!