Date: Wed, 31 Dec 2003 23:41:07 +0900 (JST) From: Hiroki Sato <hrs@FreeBSD.org> To: ue@nathan.ruhr.de Cc: hrs@FreeBSD.org Subject: Re: doc.common.mk 1.3 Message-ID: <20031231.234107.23005387.hrs@eos.ocn.ne.jp> In-Reply-To: <20031231120956.GC45118@nathan.ruhr.de> References: <20031230004119.GA19005@nathan.ruhr.de> <20031230.194507.56983609.hrs@eos.ocn.ne.jp> <20031231120956.GC45118@nathan.ruhr.de>
next in thread | previous in thread | raw e-mail | index | archive | help
----Security_Multipart(Wed_Dec_31_23_41_07_2003_337)--
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Udo Erdelhoff <ue@nathan.ruhr.de> wrote
in <20031231120956.GC45118@nathan.ruhr.de>:
ue> uh, no. My point was that a pre-defined LANGCODE is OVERWRITTEN (with
ue> a bad value) unless WWW_LANGCODE is not defined. But this is only a
ue> symptom of the real problem.
My changes in the previous mail are that make the guess of LANGCODE
and WWW_LANGCODE be performed only when both LANGCODE and
WWW_LANGCODE are undefined. I could reproduce the symptom you
pointed out:
% make LANGCODE=foo -VLANGCODE
.
but, after applying the patch it should be fixed:
% make LANGCODE=foo -VLANGCODE
foo
At least, a src/release/doc building in a non-standard directory succeeded
by specifying the correct LANGCODE and DOC_PREFIX on my FreeBSD box.
ue> The underlying question is: Why do I have to worry at all about www
ue> while working on docs? The bar for working on FreeBSD docs is already
ue> quite high with all the programs you need. Making it neccessary to
ue> think about the www directories raises this bar even higher.
No, I did not intend to make www/ tree necessary during the doc/
tree build. The problem is the LANGCODE/WWW_LANGCODE guess
algorithm's irrelevant behavior, right? With the patch, to defining
LANGCODE and/or DOC_PREFIX, a developer can work in a non-standard
doc directory, without thinking about WWW_LANGCODE and www/ tree.
Am I still missing something? If so, please show more concrete
example of the problem. Again, LANGCODE specified in the command
line is not overridden in my patch. In addition, I just improved
the guess algorithm a bit.
ue> In other words, your patch changes too much and introduces yet
ue> another set of variables that developers have to think about.
ue> And it links doc and www building together, allthough the two
ue> should be separate, clearly violating POLA.
What part that developers have to think about is changed?
Please explain. I am willing to revert my changes if the
developers feel difficult. I have added some new features,
but altogether I do not think it violates POLA.
Yes, the LANGCODE override problem is my fault, but I believe
it should be fixed in the patch previously posted. Currently
*full* www/ tree building needs the doc/ tree, but the doc/
tree building does not need the www/ tree at all (I am planning
to make mirrors.xml feature respect ${WEB_ONLY} to separate www/ tree
building from doc/ tree completely, though).
--
| Hiroki SATO
----Security_Multipart(Wed_Dec_31_23_41_07_2003_337)--
Content-Type: application/pgp-signature
Content-Transfer-Encoding: 7bit
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (FreeBSD)
iD8DBQA/8uAETyzT2CeTzy0RAvTzAKDNbs2tqZHEF+xcGXiku47oTkfvzwCeN/Uu
R29Czrj2ce0wu1BFDMAGJC0=
=d/sa
-----END PGP SIGNATURE-----
----Security_Multipart(Wed_Dec_31_23_41_07_2003_337)----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031231.234107.23005387.hrs>