In-process wineserver

[Daniel Santos]

On 01/23/2012 11:01 AM, Juan Lang wrote:
> Hi Daniel,
>
> On Mon, Jan 23, 2012 at 3:15 AM, Alexandre Julliard <julliard at winehq.org> wrote:
>> Daniel Santos <danielfsantos at att.net> writes:
>>
>>> I've updated my in-process wineserver hack, cleaned it up a bit more and
>>> fixed a few problems.  So, at least in Star Wars Battlefront II, the
>>> sound and HID problems are fixed (the "select" server call must be made
>>> via the pipe).  I presume there are other server calls with similar
>>> properties.  Also, the wineserver now exits properly and cleans up it's
>>> instance when the main process terminates.  Changed the enabling
>>> enviroment variable to "WINESPEEDHACK".
>>>
>>> So what would it take for this to be integrated into wine as an optional
>>> feature?
>> You'll probably have to kill me first ;-)
Do you have a taster? :D  (jk)
> a more serious reply:  you may not be aware the Transgaming did just
> such a thing years ago, and lobbied pretty heavily for it to be
> included.  AJ declined at the time.  The reasons for not including it
> remain the same:  because of the possibility of a buggy app taking
> down the wineserver, we couldn't ever attempt to support such a beast.
>  Any bug reported against it would be suspect.  Worse, people might
> report bugs neglecting to mention that they'd enabled it.
>
> So no, we really won't ever include such a feature as part of an
> official release around here.  We already have enough trouble with
> people reporting bugs in unsupported configurations.
>
> You're welcome to maintain your own fork, or maintain your patch for
> people to try in their own Wine builds if they like.
> --Juan
Thanks for the history and clarification.  Indeed, it's an ugly little
beast.  I've tested it on a number of apps and uncovered some other
other flaws in the patch (no surprise there).  However, I've so far
found it handy for testing some things, so I think that it's useful. 
For example, there a few performance bottlenecks I want to create
specific (conventional) patches for and I already know it will help
because I can toggle how the call is made (well yes, I know, I can also
toggle logging, and I did that before, but this just confirms it).

So I can understand your reluctance, but I still think it might be worth
consideration at some point (or in some other incarnation).

Daniel