<Pine.OSF.4.02.9907271042250.24709-100000@gromit.dlib.vt.edu>
Reply-To: buckel@wmad95.mathematik.uni-wuerzburg.de
FCC: ~/Mail/sent-mail
--text follows this line--

Dear all,

>>>>> "Paul" == Paul Mather <paul@gromit.dlib.vt.edu> writes:

    Paul> On Tue, 27 Jul 1999, Pierre BERGDOLT wrote: => I install
    Paul> netbsd 1.4 on a DEC 5000/240 a few weeks ago, and it work =>
    Paul> quite fine with twm window manager. In order to have another
    Paul> => environnement, I get the package source 'pkgsrc' and try
    Paul> to compile kde => 1.1, but it fail in compiling kdelibs. Is
    Paul> there anyone here that manage => to get kde working on his
    Paul> pmax port of NetBSD.  => Any help or suggestion is welcome.

    Paul> I have KDE 1.1 running on my DECstation 3100 under
    Paul> NetBSD/pmax 1.4.  It took a bit of "manual intervention,"

Great gosh, this should be _slow_ ;-)

    Paul> but eventually I got most of it to compile.  (I think the
    Paul> IRC client in kdenetwork is the only one I bailed on.)

I don't remember what I had to leave out in the compilation process,
but IIRC my KDE is rather complete..

    Paul> As I recall, most of the errors were attributable to the
    Paul> compiler running out of memory.  (I was surprised to see
    Paul> that the maximum amount of VM allowed a process is only ~32
    Paul> MB!)  The "fix" I adopted for this was to compile

<AOL>Me too!</AOL> At first I thought it was a bash limitation and
poked around with ulimit until I figured out that this must be kernel
related. Simply compiling the offending files without optimization
halped here.

    Paul> individually those that broke that way, but without the
    Paul> "-O2" optimisation.  (This is the "manual intervention"
    Paul> [with cut-n-paste] of which I speak.:)

Ah - I should have read this before writing. Naxt thing here were
missing moc-files which I had to rebuild manually by finding *.h-files
and moc'ing them.

    Paul> I also recall that the libtool part of the configure script
    Paul> gave problems (at least in the kde{base,network,etc.}).  I
    Paul> think what I did for this was to edit out the
    Paul> "--no-verify"(??) option to the libtool configuration in the
    Paul> configure script (so that configure would actually run to
    Paul> completion), and then let the configuration part of the make
    Paul> complete.  Then, I'd edit in the appropriate pkgtool command
    Paul> in config.status (stolen from kdelibs, I believe), and then
    Paul> re-run config.status, and then run make again.  Inelegant, I
    Paul> know, but I *do* have a running KDE... ;-)

Wow - talk about black magic here ;)

    Paul> For all I know, these problems (except for the compiler VM
    Paul> limitation) may have been eliminated in the KDE 1.1.1
    Paul> package.  (I suspect so, because I didn't have any
    Paul> *compilation* problems when trying to make the KDE 1.1.1

Well, besides of the missing moc-files there was only some minor
quirks but one should ba able to get a running KDE (if even I could
manage it, it can't be _that_ hard)...

    Paul> package on my NetBSD/alpha.  [Having something that *runs*
    Paul> is another story entirely! :-)])

<GRIN>

HTH,
			       Bernhard

-- 
      Unix is like a wigwam: No fences, no gates, apache inside.