Re: Python's deploy directory missing

Jonathan, commenting on Joel S's deploy-directory woes, says:

> Not looked at the output too closely, but I usually find that these sort of
> "hmmm, I'm sure I was already doing that" errors from ARK are a sign that
> there's a cyclic dependency.

I'd guess that is right.  (And I would wholeheartedly agree
that ARK gets really ugly when it stumbles into a cycle;
it is on my list of "infelicities".)

For a workaround, I'd suggest running *without*
--use-deps=max (I almost always just go with the default,
which is --use-deps=normal), and see what happens.

Do you have a <constraint> in your team's host/ALL.xml for
PYTHON?  i.e. not like this:

   <PYTHON>
   <string>@team:ark-dirs:LOCAL@/bin/python</string>
   </PYTHON>
   <PYTHON-BOOT>@team:ark-dirs:ARK_BOOT@/bin/python</PYTHON-BOOT>

If so, an alternative, is to *remove* the constraint and see
how you get on (with --use-deps=max, if you like).

Whatever you do, your ARK state may now be wonky (for
python), and you may need to undo some of it.  'cd
$STATE_DIR/package/<whatever> ; rm various.xml' will do
it... even rm *.xml is usually harmless.

"ARK_DEBUG=15 ark package ..." will report all the
constraint-chasing, if it comes to that.  Hope this helps,

Will

PS: I've got a plan for that rsync dependency; code to
follow...


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf