Arie Tal asked what was the status about Wine portability across non x86 platforms.

Jeremy White gave the outlook of the situation: As far as porting Wine, I think the consensus is that it's probably not worth it, due to performance issues. If anyone wishes to do this, though (especially for Alpha), Compaq has told me that they would be willing to release a free run time of their x86 emulator, so that this could be done at high speeds. Ulrich also did a good deal of work on this.

As far as Winelib, I think the belief is that porting to a non x86, Unix with X Windows should be fairly straightforward. There will be a fair amount of work to find and remove all of the unknown bugs with byte endianness and packing, but the fundamental design and plan for Wine is to make this easy.

For non Unix and/or non X Windows systems (such as BeOS and MacOS), it gets a little bit harder. BeOS seems to be much harder because it is missing many systems calls that Wine relies upon (Patrik is the expert on Be, and I believe there is a web page dedicated to Wine on BeOS). MacOS <= 9 is similarly hard.

MacOS X, however, is another story. If you get an X server on MacOS X, it should be pretty easy. If you want to do it the right way, you'd need to develop a Carbon driver to parallel the Wine x11 driver. That, while conceptually straightforward, will be a lot of work.

Gavriel State reported: I got some simple winelib apps up and running on LinuxPPC in early 1999, but it was fairly hackish, and most of my changes didn't make it into the mainline WINE tree. I did some work updating the port for newer releases of LinuxPPC at MacHack this summer, but got stuck trying to figure out how to properly deal with the thread local storage. My old code still works ok though, if you happen to have a LinuxPPC machine around.

Ulrich Weigand spoke the last word, reporting he had Winelib working on Sparc: The main Wine tree doesn't compile out of the box. There's still a few problems, mostly related to alignment issues, that I haven't yet fixed in a clean way.

On machines without alignment constraints (even if big-endian), there shouldn't be much changes necessary. For example, just for fun I tried to get Wine to compile on S/390; it took me only a few hours until WineMine was running ...

but who would need winmine on S/390 ???

Jonathon Griffiths, continuing his CRTDLL effort (please refer to previous article), asked a few more questions:

• Locking: at some point I will need to add locking code for MT safety. Are there any issues (performance?) I need to be aware of using the win32 locking functions in the dll?
• Thread-safe: MS docs state that crtdll is not very thread-safe, however I can't see any reason why wines shouldn't be. I don't think any apps depend on non thread-safe behavior since by definition it is unpredictable. So I think it may be worthwhile to make it MT safe. This would mean a large part of the code could be shared with msvcrt.dll (way in the future).
• #includes: Wine's "process.h" will conflict with the crt "process.h". Is it desirable to have crt headers in the wine include dir, or should they be in include/crt? (note I wont be writing headers for some time).

On the #include part, everyone agreed that Wine's internal process.h include file should be moved to Wine only include directories, and a conformant process.h should be created in the include directory.

On the thread safeness issue, Alexandre Julliard was strongly against making CRTDLL thread safe: it isn't under Windows, so it shouldn't be under Wine. Gav State, however, did like the idea of using CRTDLL but still having thread safeness. Alexandre wasn't still convinced, and preferred to have Wine also to provide a native msvcrt.dll (which would be thread safe).

Douglas Ridgway announced: We're going to try to cut over the mailing lists from the old server to the new server today. Hopefully, we will be able to not lose subscription info etc. The current plan is to change over to using majordomo at the same time -- I'll provide some details once it's done.

Even if some people objected the use of majordomo and preferred mailman, the move has taken cleanly (YMMV), and the good part of it is that the NNTP access to wine-devel, wine-announce, wine-patches and wine-cvs is back online.

Ove Kåven, while browsing the Debian Wine bug database, ran into a user's proposal which pleased him well. Lots of users had a hard time configuring Wine; Wine installs (by default) its .so files in /usr/lib/wine, which requires that either /etc/ld.conf or LD_LIBRARY_PATH to list this directory in the to search for list. However, some RPM installations (and lots of by hand installations) forgot to do so, leading to numerous but irritating questions on comp.emulators.ms-windows.wine newsgroup.

The main idea of the patch is to make use in the .so file link option of the -rpath option, which tells where to look for the .so file. Currently, the build options were pointing to /usr/lib instead of /usr/lib/wine.

However, Ove also wanted to be able to drive this directory with a --dlldir option in configure. Steve Langasek proposed a quick way of doing it.

Gérard Patel also expressed his will to still be able to keep several Wine binary trees in parallel (for testing reasons). Ove didn't think this should be too much of a problem, even if the details haven't sorted out yet.

This small patch should definitively help in installing Wine.