Re: [AD] [DOS] Win2k mouse bug

["Eric Botcazou" <ebotcazou@xxxxxxxxxx>]

Henrik wrote:
> allegro successfully detects if NT is installed by checking if the
> environment variable OS == "Windows_NT". Can you check what int 0x2f,
> ax=0x160a returns in ax, bx and cx for windows nt 4 and windows 2000?
> that's the only thing i can think of for the moment. this works at least
> on windows 3.1 and up to windows ME. for reference check ralph brown's
> interrupt list.

Eli Zaretskii told me the guys from comp.os.msdos.djgpp had already tried
all the known methods and the results are strictly the same under WinNT4 and
Win2k.
With Vincent we checked that Win2k returns 0x532 (v5.5) for
_get_dos_version(1) and "Windows_NT" for getenv("OS").
I also tried some trickery with int33h itself, but in vain.

> there should also be a somewhat undocumented VxD call in NTKERN for
> returning NT kernel version. (int 0x20, vxd=0x004b)
>
> should i code a test program?

Please do and send it (in compiled form) to Vincent.


Laurence wrote:
> Here are a couple:
>
> - examine the output of the `ver' command

Same result: v5.5 for both.

> - look at some file which is present in both OSes but differs in
>   filesize or contents (eg. $WINDIR/win.com or some such thing).

Is this really reliable ? You will have to take care of various Win2k
versions, localized versions, service packs...


Grzegorz wrote:
> Maybe a force configuration option in allegro.cfg? I think that's the most
> reasonable for the moment, until there's some way to find out if W2k is
> there (maybe searching for critical files in windows directory?).

I don't see any other solution.


Javier wrote:
> I would say right now there are more chances to see an allegro user using
> Win2K than WinNT (and in general, games). So i would go with Win2K
> support...

Same story as with DirectX 3 vs DirectX 5: should we drop WinNT4 support ?


Chris Jones:
> How about detecting for the presence of the LFN API, which Win2k supports
> but NT4 doesn't?

Are you sure ? I would say that the djgpp guys have already tried the LFN
approach, but could you code something though ?

--
Eric Botcazou
ebotcazou@xxxxxxxxxx