From: Paul K. <pki...@us...> - 2003-03-01 16:33:04
|
Andy Adler wrote: >On Fri, 28 Feb 2003, Paul Kienzle wrote: > >>Andy Adler wrote: >> >>>Paul, >>> >>>I'd like to upload my octave-for-windows to octave-forge. >>>I've not had a bug report for a few weeks now. >>> >>>Could you create a "octave for windows" package >>>category and I'll add the file there. >>> >>> >>> >>Don't we already have one in the form of octave-forge-cygwin? Or is >>this a pure octave package? >> >> >This would be for self contained octave-for-windows, using >the nsis installer. > >I consider this to be distinct from octave-forge-cygwin, >which I intend to keep up, as well. > The difference is that octave-forge-cygwin is a developer tarball which installs into an existing cygwin installation and the other is a stand-alone which does not require cygwin? Ideally, I would like the developer version to be tied to the standalone version so that oct-files created with the developer version can be installed in the standalone version and just work. In that case, I would like the user and developer versions to be different release files in the same release. I know sourceforge will support this since I've seen it for example in the tcl file release page: http://sourceforge.net/project/showfiles.php?group_id=10894 I realize I've said all this before and we are not there yet. Some issues that remain to be worked out: * do we ship multiple dll's or one big exe which we link octfiles against? * do we build with a light kernel or do we build all the DLD-FUNCTIONS as builtins? * if heavy kernel, do we build octave-forge as oct-files or do we make a super heavy kernel which includes them all? * can we build lapack-plain.dll and various lapack-atlas dlls so that atlas enabling is just a matter of copying the proper dll to lapack.dll? * do we build libstdc++.dll? Or do we dump the contents of libstdc++.a into libcruft or (in the heavy kernel world) into the octave.exe? I've tested the light kernel option. It works fine except that the fork problem on recent cygwin for Win 9x is exacerbated (not only is it slow, it also pops up error messages for each loaded oct-file). The solution is to purge the octave code of fork, but I haven't had a chance to do so yet. The heavy kernel option will need some work to make sure it can support the separately compiled option. If super heavy is the preferred option, we could rework the octave-forge cc files so that they use the DLD-FUNCTIONS style of #include "blah.h" rather than #include <octave/blah.h>. That way you don't have to construct the octave include directory as part of the octave source tree. Then it is only a matter of changing octave/src Makefile so that it automatically builds everything in the DLD-FUNCTIONS directory which means that we don't have to edit anything by hand when building a new release version. Comments? Paul Kienzle pki...@us... |