Ivans comment that the Mandrake RPMs didn't work on Red Hat/Fedora surprised me, so after encountering a stuck Mandrake user for whom the RPMs weren't working either on IRC I decided to check them out.

It turns out the reason they don't work on Fedora is because they don't work at all - wine.inf is in /usr/share/doc/wine-20040615 which is wrong, it needs to be in /usr/share/wine. As this file is in the wrong place Wine refuses to start on a new users account.

Once I manually copied the file to the right place Wine worked just fine.

It seems that we are getting ever so closer to 0.9. Slow, but steady. Looking at the TODO, it looks like upgrading is one of the big showstoppers, in that it will affect (1) the end-user experience, and (2) how we deal with the configuration.

So this seems like a good time to start discussing this topic, as we need to eventually reach the elusive 0.9. This seems to be a difficult topic, so we need to approach the problem well prepared. That is, we need to first define (and agree) on what we need to solve here. So after a tumultuous discussion on IRC, I decided to get the ball rolling.

Intuitively, upgrading wine is simple to understand: once a new version is installed, we need to get users in a state where they can use it. While simple to state, this problem is complicated by the various corner cases that can appear in Real Life (TM):

• native vs. builtin DLLs handling
• multiple Wine installations on a box
• updating only some DLLs
• integrating into the Unix environment
• dealing with Windows software

We can also look at some important use-cases that we need to support, but before we do, the following must hold true in all cases:

• administrators should be able to install a new Wine on the system easily with "rpm -U" (or equivalent).
• the upgrade should not erase registry settings that are managed by the user

Let's look at the use-cases:

A. Global Install

In this case, both Wine code, as well as applications are installed globally, for all users to access. This is the most Unix-like setup, but unfortunately the most difficult to support, due to the fact that Win32 apps simply expect a different environment. Since we can't control application's code, this setup may present some limitations, but it may be sufficient for sites that need a limited set of applications. For this to work, it is acceptable that we ship special scripts that know about the applications that are supported, so that they can work around inherent limitations in the applications themselves. Users should transparently be able to use the new version of wine, the next time they start wine.

B. Mixed Install

Here, Wine code is installed globally, but applications are installed in the user's account. In a way is like (A) but with some additional applications installed directly into the user's account. Same requirements apply to this case as well.

C. Local Install

Both Wine code, as well as applications are installed locally, into the user's account. The Wine code is still installed globally by the administrator (as in A & B), but it is copied locally before it's being used. An upgrade in this case may require an explicit action by the user, so the upgrade can happen at a controlled time, when the user desires it.

In all of the above, there is always state in the user directory. Thus, it is imperative that the upgrade process makes sure that after the upgrade, the user-private state is consistent with the new code base. Also, any solution must take into account that the registry must be created on the fly, as it is created when registering the DLLs.

Before we go any further, I'd like to ask everybody to contribute their thoughts/requirements/desires so that we get a conprehensive view of the problem we are trying to solve.

I am currently working on the janitorial task of getting rid of W->A calls and I have a few questions about how I should implement these changes.

I'll start the question with an example:

dlls/advapi32/crypt.c line 255, 482

Initially CryptAcquireContextW converts all of its arguments to ansi then calls CryptAcquireContextA (which is intuitively not what we're wanting and thus the cleanup.)

The question is, what exactly is the best way to clean this up?

It seems to me that there are two main options as to what can be done:

A) convert the W->A calls in the wide-char functions to A->W calls in the asci functions and implement the wide-char functions instead, or

B) implement both asci and wide-char functions separately ie asci functions only call asci functions etc.

If we are to choose plan A, I have seen several different methods of converting asci to wide-char, and I am wondering which method is best.

I would have to argue that implementing both the asci functions and wide-char functions separately would be the most efficient way to solve the W->A problem. Throughout all the wine libs, many conversion calls are being made from W->A and A->W, when most of these calls could be left out. I think performance would improve without all of the conversion calls. Granted the size of the code would increase, but speed would also increase.

There are a couple snags when it comes to implementing asci and wide-char functions independently, but that can be discussed later.

If I'm totally off the mark, let me know, and if there are other jobs I could be doing, let me know as well because I would like to contribute, but I'm not exactly sure what to do yet. Thankyou for your time.

I did not check your code, but each situation has its reason and you decide which you will be better. I suggest you to start converting something simple, like:

dlls/shell32/systray.c: shell32: Shell_NotifyIconW: illegal call to Shell_NotifyIconA
dlls/version/install.c: version: VerInstallFileW: illegal call to VerInstallFileA

When I made one of these conversions, I chose "dlls/shell32/shlexec.c: shell32: ShellExecuteExW: illegal call to ShellExecuteExAand" very laborious and was complicated. You can give one looked at as I made: http://cvs.winehq.com/patch.py?id=10692