Excessive dependancies for OpenOffice 2.0 port

[Volker Quetschke]

Mikhail T. wrote:
> => Actually, I do realise that, and, in my opinion, it is _harder_
> => to keep it building the current way. The third-party packages --
> => installed by other ports -- have their own maintainers, who watch out
> => for build problems.
> 
> =Then I suggest that you provide the patches and feed them back to OOo
> =so that the next time you don't have to do it again once a new OOo
> =version comes up. Don't forget to sign the copyright agreement.
> 
> I am doing that, but I'm NOT an OOo developer. My patches tend to be
> FreeBSD-specific. FYI, I too have some experience in porting and I am
> well aware of the benefits of the vendor's accepting the patches.
> 
> "Feeding them back" to OOo is a good idea, but it should not be
> holding back the progress of the FreeBSD port. And that is my point.
> In "lawyers' speak", the goals of the OOo and the FreeBSD are similar,
> but not identical. Maho has a "conflict of interest" -- and I feel the
> FreeBSD port is under-represented.

IMHO there is no conflict of interest. (I shouldn't speak for Maho here,
sorry for that.) There /should/ be no conflict as we all should only
want a working and easily buildable OOo. And "under-representation" is
obviously not solved by working on ports and not contributing back the
patches to OOo. The mysterious "someone" that represents the evil OOo
can only be Maho in this case.

> And, obviously, I don't care for a "copyright agreement". The patches
> are/will be there for all to see, and -- by virtue of being part of a
> FreeBSD port, they will, of course, be BSD-licensed.
That you feel that way about the copyright assignment will only make it
harder for "them" to include your patches.

> => Building a special version of C compiler is, AFAIK, unprecedented.
> =Feel free to work around the bugs that prohibit the use of *old* gcc
> 
> I do feel very free, thank you. But you are not really countering my
> point here... Requiring a port-specific C-compiler is unprecedented. You
> stated the problem, but the lang/ooo-gcc is NOT a solution to it.
Well, for a lot of linux distributions it was a solution to apply the patch
to their system gcc. I assume the current FreeBSD system gcc is free of
additional patches, so that would be unprecedented. ;)

> =Even though you have your point that a lot of packages are included as old
> =versions there is also work going on to reduce these dependencies. Did
> =you ever look at the "--with-system-XXX" parameters of OOo's configure?
> 
> Yes, OF COURSE, I did "ever look". Very nice of OOo to finally wise up
> to this (1.x releases had very little flexibility in this). Too bad,
> FreeBSD port does not use this, even where it can.
As Maho said, there are problems (Bugs) with some of them. Maybe worth
fixing?

> =Also you complain about a tool (dmake) that takes roughly 10 seconds
> =to build *and* that is available in the ports collection so that it
> =doesn't have to be rebuild? Do you have the time to check/redo all the
> =makefiles for a 300MB source package that builds just fine with dmake?
> 
> I don't have the time (to re-write makefiles into BSD or GNU make), but
> I don't object to using dmake per se. I object to using the bundled one
> instead of the devel/dmake.
Hmm, the OOo configure checks for an installed version of dmake, yes, platform
independent, if it doesn't use the installed one and builds the included
version this is a bug. But maybe just another dependency is needed.
(Didn't check)

> =Maybe you should check your attitude. If you stop thinking of OOo as a
> =port but as an application that runs on many OSs (including FreeBSD)
> =and the failure to do so as a bug we might actually get somewhere.
> 
> I sure appreciate the OOo project and Maho's contribution to it.
> But I am talking about a different thing called "FreeBSD port
> editors/openoffice.org-2.0". Maho -- and, perhaps, yourself -- are OOo
> developers (nothing wrong with that). I'd like a good FreeBSD port...
Agreed :)

> =I see that you are already familiar with the internals so that I don't
> =have to point you to Jan Holesovskys (and others) work on the 64bit
> =version.
> 
> Thanks, I'm aware of Kendy's work and have exchanged e-mails with him
> already as well. Jan's work for the Debian port is an example, of what
> is lacking in the FreeBSD port. He even has patches, that HE KNOWS will
> never be accepted "upstream". Imagine that!
Yes, I don't know the specific patches here, but I can imagine that tons
of

#if defined(somemacro)
   blah
#else
   blubb
#endif

constructs are likely to be rejected. It is a maintainance nightmare.
And it doesn't matter if somemacro is related to 64bit or MACOSX, FREEBSD,
WINDOWS, SOLARIS, you name it.

Unfortunately it is a lot harder to write a nicely mainainable solution
than the easy ifdef sprinkling. I also don't say that the code as is is
perfect (it already has a lot of #ifdef sprinkling) but the number of
supported OSs has grown and so have the expectations of the code quality.

Volker

-- 
PGP/GPG key  (ID: 0x9F8A785D)  available  from  wwwkeys.de.pgp.net
key-fingerprint 550D F17E B082 A3E9 F913  9E53 3D35 C9BA 9F8A 785D
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20051107/13c36e10/signature.bin