[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: SWIG Again

From: John Lenz <lenz_at_cs.wisc.edu>
Date: 2004-12-16 22:26:47 CET

On 12/16/04 10:35:28, Philip Martin wrote:
> John Lenz <lenz@cs.wisc.edu> writes:
> > Using this patch, you SHOULD be able to download 1.3.24, extract it in
> the
> > subversion directory, and rename SWIG-1.3.24 to SWIG.
> I prefer to use an installed SWIG, I assume that will still be
> supported.

Yes, if you look in swig.m4, it first checks the installed SWIG version,
and if there is no installed swig or swig is not recent enough, it then
checks inside the build directory.

> > A few problems...
> > 1) It doesn't seem to want to check the version correctly.
> That sounds like a potential show-stopper, what goes wrong?

The problem is in that SWIG_BUILD_VERSION thing below, sed exits with an
error. I just need to figure out what sed doesn't like.

> > + # We want the version as an integer so we can test against
> > + # which version we're using. SWIG doesn't provide this
> > + # to us so we have to come up with it on our own.
> > + # The major is passed straight through,
> > + # the minor is zero padded to two places,
> > + # and the patch level is zero padded to three places.
> > + # e.g. 1.3.21 becomes 103021
> > + SWIG_BUILD_VERSION="sed -e 's/[[^0-9\.]].*$//' \
> > + -e 's/\.\([[0-9]]\)$/.0\1/' \
> > + -e 's/\.\([[0-9]][[0-9]]\)$/.0\1/' \
> > + -e 's/\.\([[0-9]]\)\./0\1/; s/\.//g;'"
> I realise this is derived from previous code, but given SWIG's past
> performance on API/ABI compatibility it may be simpler just to match a
> list of known working versions, at present that just "1.3.24".

I keep trying to repeat this, but no one seems to get it. So you will have
to pardon my shouting.
let me repeat that

If you generate a file from swig, it DOES NOT need to link with anything AT
ALL from swig. Just forget everything you ever knew about libswigpy. It
is gone. Two wrappers generated by swig DO NOT need to link with each
other either.

Yes, SWIG has had a previous history of ABI problems, mainly because 1.3 is
still a development branch and we want to change stuff around. When 2.0
comes out, then this stuff won't happen. It actually doesn't matter
anymore now because there is no more ABI. I would say SWIG solved its
history of having ABI problems by just removing it. (Note, no
functionality was actually removed, the stuff done by libswigpy is now
accomplished directly by the wrapper).


To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Dec 16 22:29:16 2004

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.