[developers] Pet dependencies
Eric Nichols
eric-n at is.naist.jp
Sat Jun 18 07:38:34 CEST 2005
Greetings,
I got fed up with hacking makefiles to solve these problems when I was
installing the latest PET on a few machines, so I put together a patch
that addresses the rmrs.h distribution issue and the renamed lib/linux
directories problem. Basically, I changed all instances of rmrs.h to
mrs.h and forced it to compile mrs.h before cheap when the MRS option
is selected. I solved the library problem by borrowing some code from
the LKB installer; now the configure process uses uname to check what
kind of linux is running and uses a variable to select the proper
directory. I have verified that this setup works on a Pentium 4 running
Fedora Core 4 with a 2.6.11smp kernel and on a Pentium M running Ubuntu
5.04 with a 2.6.10 kernel. I have not been able to test this on any 64
bit machines. Any testing and/or feeback would be greatly appreciated. I
also verified that PET 0.99.7 does not compile using the new 0.9f
release of ECL. Maybe we should look into this in the near future.
On 5/4/05, Francis Bond <fcbond at gmail.com> wrote:
G'day,
I tried to install pet-0.99.7 from scratch, using a recent CVS lkb,
and ran into the problem Stephan mentioned with lkb/include/mrs.h no
longer being distributed.
Therefore, if you want to use the mrs machinery, a possible solution
is to:
(1) configure once (with eclmrs set but not recognized)
- lkb/include/rmrs.h does not exist
(2) go into the cheap directory and run "make libmrs.a"
- this makes delphin/lkb/include/mrs.h
(3) ln -s delphin/lkb/include/mrs.h delphin/lkb/include/rmrs.h
(4) configure again
(5) make
A better solution would be to:
(1) remove the check for rmrs.h in the configure
(2) make sure make libmrs.a gets called early on, and
either: link the resulting mrs.h to rmrs.h
or: harmonize the names
I am afraid my configure writing skills are not even up to this simple
task, but I would be happy to test things for people.
On the subject of harmonization, Stephan but in a renewes plea for the
following:
> while we are at it, francis documented (for the `main' branch):
> ./configure --with-tsdb=... --with-eclmrs=... --with-icu=...
> any chance we could still harmonize that to, --with-lkb and --with-ecl?
> for the first one, we are picking up the MRS library and [incr tsdb()]
> libraries from the LKB tree. i am currently experimenting with adding
> the finite-state preprocessor to that collection, hence `--with-tsdb'
> really is a bit of a mis-nomer. for the second path, i take this to
> point to the ECL root directory (i.e. `--prefix' when installing ECL),
> at least according to what francis put on the PET wiki? in that case,
> `--with-eclmrs' is not quite the right name either, in my view.
On a related note, cheap looks for libitsdb.so in delphin/lib/linux
and these files are now in delphin/lib/linux.x86.32 or
delphin/lib/linux.x86.64. I just linked them on my laptop, but there
is probably a better solution.
Thanks,
--
Francis Bond < www.kecl.ntt.co.jp/icl/mtg/members/bond/>
NTT Communication Science Laboratories | Machine Translation
Research Group
--
Eric Nichols < ericnichols79 at gmail.com>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pet-0.99.7.patch
Type: text/x-patch
Size: 7389 bytes
Desc: not available
URL: <http://lists.delph-in.net/archives/developers/attachments/20050618/734861f7/attachment.bin>
More information about the developers
mailing list