Re: updating ConTeXt with ctxtools

[Hans Hagen <pragma@wxs.nl>]

Sanjoy Mahajan wrote:
>> In any case, I have probably misunderstood the thread. Has a summary
>> been posted somewhere that I might consult?
>>     
>
> I need to summarize all the Linux installation methods on the Wiki,
> once I get it all straight in my head.  Right now in reality and in my
> head it's a nightmare because of all the flavors (texlive, tetex 2 or
> 3, etc.) and the gazillion map files that could be anywhere.  ConTeXt
> tries to work everywhere so it has a hard life.
>
> I don't know much about what MacOS X does, but if you're not using a
> graphical installation method, it should be like BSD 4.[0123] Unix
> that I once used.  In which case it should work mostly like Linux.
>
> Here's what I do now and did just half hour ago to update to the
> 2006.08.08 version.  I use bash (a mostly sh-compatiable shell), which
> the MacOS terminal runs too I think.  The $ below is my bash prompt
> and everything after it is what I type:
>
>   $ TEXMFLOCAL=/home/sanjoy/texmf texmfstart ctxtools --updatecontext
>
> This fixes the problem that ctxtools uses (I think) Windows quoting in
> this source line:
>
>             tree = `kpsewhich --expand-path $TEXMFLOCAL`.chomp rescue nil
>
> In Unix ruby (like in Perl) the $TEXMFLOCAL is expanded by the shell
> because of the $, and the result is run by the backticks.  Usually
> TEXMFLOCAL has no setting in the shell or environment so ruby runs the
> command with $TEXMFLOCAL replaced by an empty string, giving:
>
>    kpsewhich --expand-path 
>
> So kpsewhich complains that it didn't get an argument for
> --expand-path (or --expand-var).  The long-term fix is to figure out
> how to quote that works in Windows and Unix, but I don't know how
> since I don't use Windows.
>
> The short-term hack is to set TEXMFLOCAL just while running ctxtools,
> which is done by the command above that I run.
>
> So maybe try that using the vanilla distribution (so with \012
> restored) and report back what happens and let's see if we can find a
> fix if something else doesn't work.  My computers run Linux but my
> partner has a MacOS 10.3 machine and I'll use it for ConTeXt
> experiments if needed.
>   

does this one work ok?

(the stupid thing is that kpsewhich is not made for user friendlyness, 
e.g.\ a natural solution would be to support

kpsewhich --expand-path TEXMFLOCAL

i.e. it expects an internal var so why not be clever and put the $ in 
front of it internally; there are more places in kpsewhich where robust 
cross-platformness can be provided but it has a low priority i guess


        def locatedlocaltree
            if System.unix? then
                tree = `kpsewhich --expand-path '$TEXMFLOCAL'`.chomp 
rescue nil
                unless tree && FileTest.directory?(tree) then
                    tree = `kpsewhich --expand-path '$TEXMF'`.chomp 
rescue nil
                end
            else
                tree = `kpsewhich --expand-path $TEXMFLOCAL`.chomp 
rescue nil
                unless tree && FileTest.directory?(tree) then
                    tree = `kpsewhich --expand-path $TEXMF`.chomp rescue nil
                end
            end
            return tree
        end


-- 

-----------------------------------------------------------------
                                          Hans Hagen | PRAGMA ADE
              Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
     tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com
                                             | www.pragma-pod.nl
-----------------------------------------------------------------