On Sat, 2005-10-29 at 19:15 +1000, skaller wrote:
> On Sat, 2005-10-29 at 18:17 +1000, Erik de Castro Lopo wrote:
> > IMO, having the pre-releases with the same name as the final
> > is a bad idea. My prereleases go libX-A.B.Crc1.tar.gz,
> > libX-A.B.Crc2.tar.gz and so on, which becomes libX-A.B.C.tar.gz
> > on release. Therefore no confusing rc with final.
> Ok, done: now Morphlix versions are like:
> and final versions are like
I am about to 'release' the flx_1.1.1_src.tgz tarball *as is*.
This is premature, but I'm releasing it to the Godi packaging
system whose last version is rather old -- 1.0.26 :)
The problem is I can't release 1.1.0 because it doesn't build.
However once released I cannot modify 1.1.1 because the checksums
Therefore, the version number is about to be bumped to 1.1.2,
the first pre-release Morphlix will be
There will be no 1.1.2 until it is stable (that is, the Morphlix
_rc<v> sequence is terminated by a decision to release).
The actual release will be the last of the rc<v> series plus
last minute patches (which of course will probably have bugs :)
I am thinking to put symbolic links on the web site so you
can get the 'latest' stable Felix or unstable Morphlix release
candidate without knowing their exact numbers.
Mike: 1.1.1 should probably be wrapped for Debian,
I am not sure if it should depend on Ocaml 3.08 or 3.09.
There is a known issue with .sonames: the current build
does NOT put a .so name in the flxlib_dynamic.so file.
The installed system just requires that library, unversioned.
I think this is justified by the following viewpoint: the
Debian packaging of Felix is NOT intended at the point to
provide the full toolkit. Instead, it is intended to
provide just a single command: 'flx' (the bash script).
This script should automatically rebuild everything
when versions upgrade, that is, any binaries generated
are regarded as 'cached' or like 'pre-compiled headers',
not final products: felix program is defined in terms
of the source, so the flx script is responsible for
ensuring the source is correctly executed, not the
This basically defers packaging the toolkit properly
as separate packages.
Please let me know if you think this is OK for now.
I would like to push some versions of Felix into the
Debian archive to get it out there, so we can get a
few more users to gain some more experience about how
to package it best -- this depends partly on user needs
and there just aren't enough users at the moment to
know what they are. Chicken and egg.. :)
I also rely on Debian autobuilder to find problems on
architectures I don't have access to.
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net