[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

Post by Thomas Loc » Thu, 10 May 2007 07:04:49


> Thanks for reporting this, I have fixed this. I have fixed this by just

OK, but that is just the first part of my problem it turns out...


I can get the missing file name into Makefile.am by hand or by rerunning
mwc.pl. But autoreconf fails to rebuild TAO/tao/Makefile.in (among
others) based on the contents of Makefile.am. I've tried mwc.pl command
line arguments which worked in the 1.5.0 days and also tried arguments
culled from the make_release.py script.

Hmm, this turns out to not be an MPC issue most likely; see below...

>> ...

For 1.5.0 I did not need the -type argument and used "TAO_ACE.mwc" as
the configuration file name (different than above). The arguments I
gleaned from make_release.py look like

/usr/src/redhat/BUILD/ACE_wrappers/bin/mwc.pl -type gnuace
-exclude 'TAO/TAO_*.mwc,TAO/CIAO/CIAO_*.mwc'
-recurse -hierarchy
-relative ACE_ROOT=/usr/src/redhat/BUILD/ACE_wrappers
-relative TAO_ROOT=/usr/src/redhat/BUILD/ACE_wrappers/TAO
-relative CIAO_ROOT=/usr/src/redhat/BUILD/ACE_wrappers/TAO/CIAO

(note line wrapping)

But in the first case there are complaints from automake that
BUILD_EXCEPTION is not a defined macro (I see it defined in m4 files in
the ACE/TAO tree). And in the second case from make_release.py I do not
seem to get a new version of Makefile.am at all!

Ah, I think I see the problem, or at least an aspect of the problem.
1.5.8 seems to have more configure.ac files than in the past after
running autoreconf, but some of the configuration files do not have
AC_CONFIG_MACRO_DIR() to help automake/autoconf locate the macro
directories. This might work if you were running aclocal manually with
some command line arguments, but I would guess that automake needs some
hints on where to find macros, or something to keep new configure.ac
files from appearing in new places.

So, is there a strategy for how to rebuild the autoconf files? What
works for 1.5.1 does not work for 1.5.8. With a bit of a hint on what is
expected I could probably generate some patches if necessary.

- Tom

--
Thomas Lockhart
Supervisor, Realtime Software Group
Interferometry and Large Optical Systems
Caltech/JPL
 
 
 

[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

Post by Johnny Wil » Thu, 10 May 2007 15:19:45

i Thomas,

The -type is now required by MPC. The autoconf support has been improved a
lot by JT as you know and he did regenerate things for us. We haven't
heard from JT for months so no work has been done on autoconf support. I
do see your point that there are issues, but I just don't have time to
invest in fixing or investigating all autoconf issues without any
commercial support agreement. If someone would fund fixing all issues, we
are happy to do it, but until that moment please deliver patches to
improve the autoconf support.

Regards,

Johnny




 
 
 

[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

Post by Thomas Loc » Fri, 11 May 2007 03:56:20

> The -type is now required by MPC. The autoconf support has been

Yes, I had gotten my initial instructions for how to regenerate the
configuration support from JT back in the 1.5.0 days, just before
autoconf support became solid for 1.5.1. I assume that "automake" is the
correct type argument for MPC? For an ACE+TAO tarball, is there a
particular choice of mwc files to use when running mwc.pl? I gave an
example I had extracted from make_release.py; does that reflect what
would be done for the ACE+TAO packaging as well as the full ACE+TAO+CIAO
package?


OK, that is my hope :) I have good understanding of traditional
autoconf, much less so of libtools (but that does not seem to be
involved here). MPC is a mystery, though again that does not seem to be
implicated in this particular problem. I'll probably need to learn more
about other aspects of automake to figure out why those new configure.ac
files are being generated.

I'll start tracking down the initial symptoms and will keep asking
questions. I'd really like to see autoconf continue to be supported
since it does make for a clean configuration and installation process
which has been a traditional weakness in ACE/TAO.

- Tom

--
Thomas Lockhart
Supervisor, Realtime Software Group
Interferometry and Large Optical Systems
Caltech/JPL
 
 
 

[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

Post by Johnny Wil » Fri, 11 May 2007 03:58:59

Hi,


Yes, it is automake. I think mwc.pl is run from the root on everything


We also would like to see that autoconf support gets improved again, if you
have patches we are happy to apply them.

Regards,

Johnny Willemsen
Remedy IT
Postbus 101
2650 AC Berkel en Rodenrijs
The Netherlands
www.theaceorb.nl / www.remedy.nl

*** Integrated compile and test statistics see http://www.yqcomputer.com/ ***
*** Commercial service and support for ACE/TAO/CIAO ***
*** See http://www.yqcomputer.com/ ***
 
 
 

[tao-bugs] [autoconf build]: TAO-1.5.8 missing file from MPC/autoconf installation

Post by Thomas Loc » Fri, 11 May 2007 09:21:28

> We also would like to see that autoconf support gets improved again,

Hmm. I'm gathering that perhaps the .am files were never completely
generated from MPC? Or at least not in recent history? I see some
comments in the changelog for TAO that indicate that this might have
been the case between 1.5.0 and 1.5.1 (where I first notice troubles
running autoreconf after running mwc.pl, though just running autoreconf
on 1.5.1 does seem to work).
Are you aware of any time after that when .am files were committed from
a clean run of mwc? If not, perhaps the best strategy is to try to
manually patch the .am files which are now broken (I see some things
which might help them, but haven't yet tried to fix them), as opposed to
trying to get the MPC templates and configuration files to work without
extra intervention. Any thoughts on this? Thanks for responding!

- Tom

--
Thomas Lockhart
Supervisor, Realtime Software Group
Interferometry and Large Optical Systems
Caltech/JPL