Technical discussion of toolchain issues

Alan Barrett | 6 Sep 2007 10:32

Re: bind, postfix, ntp break MKDOC=no?

On Thu, 06 Sep 2007, Thor Lancelot Simon wrote:
> Is it expected that MKDOC=no will cause build.sh to fail at the 'sets'
> stage?  This looks like a whole lot of precedent...
> 
> =======  453 extra files in DESTDIR  =========
> Files in DESTDIR but missing from flist.
> File is obsolete or flist is out of date ?
> ------------------------------------------
> ./usr/share/doc/html/bind9/arm/Bv9ARM.ch01.html
> ./usr/share/doc/html/bind9/arm/Bv9ARM.ch02.html
> ./usr/share/doc/html/bind9/arm/Bv9ARM.ch03.html
[...]

It looks as though the lists are correct (those files are tagged with
the "doc" keyword, which means taht they should not be prefent when
MKDOC=no).  src/usr.sbin/bind/html/Makefile also appears as if it should
do nothing if MKDOC=no.  I haven't checked the other (non-bind-related)
files in your list.

I suspect that you had those files lying around in your DESTDIR, as a
result of an earlier build without MKDOC=no.  if you use the build.sh
"-u" option, then any options that affect which files are produced by a
build must be kept consistent from one build.sh run to the next.

--apb (Alan Barrett)

headers

Thor Lancelot Simon | 6 Sep 2007 12:09

Re: bind, postfix, ntp break MKDOC=no?

On Thu, Sep 06, 2007 at 10:32:48AM +0200, Alan Barrett wrote:
> 
> I suspect that you had those files lying around in your DESTDIR, as a
> result of an earlier build without MKDOC=no.  if you use the build.sh
> "-u" option, then any options that affect which files are produced by a
> build must be kept consistent from one build.sh run to the next.

Yes, I'm aware of that.  The section of log I sent was from a build.sh
invocation without -u but with -V MKDOC=no, unfortunately.

--

-- 
  Thor Lancelot Simon	                                     tls <at> rek.tjls.com

  "The inconsistency is startling, though admittedly, if consistency is to
   be abandoned or transcended, there is no problem."	      - Noam Chomsky

headers

Alan Barrett | 10 Sep 2007 09:36

Re: bind, postfix, ntp break MKDOC=no?

On Thu, 06 Sep 2007, Thor Lancelot Simon wrote:
> The section of log I sent was from a build.sh
> invocation without -u but with -V MKDOC=no, unfortunately.

I have successfully performed a build.sh with the following options:

./build.sh -T ${TOOLDIR} -O ${OBJDIR} -D ${DESTDIR} -R ${RELEASEDIR} \
	-u -U -V MKMAINTAINERTOOLS=yes -V MKDOC=no tools release

Both host and target were NetBSD/i386 4.99.30.

There do seem to be some issues with "-j", which I haven't yet
investigated in detail.

--apb (Alan Barrett)

headers

sosoysq | 11 Sep 2007 02:47

remove

remove