[Bug 56225] New: Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Bug ID: 56225
           Summary: Myst deadlocks on start since Wine 3.2
           Product: Wine
           Version: 3.2
          Hardware: x86
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Steps to reproduce:

1. Run `sudo sysctl vm.mmap_min_addr=0`

2. Run `DISPLAY=:1 wine MYST.EXE`

In Wine 3.1, the game pops up a dialog that says "Myst requires a 256-color
palettized display driver." But in Wine 3.2 and later, Wine just prints an
error like the following:

00f0:err:sync:RtlpWaitForCriticalSection section 777129E0
"../wine/dlls/krnl386.exe16/syslevel.c: Win16Mutex" wait timed out in thread
00f0, blocked by 0114, retrying (60 sec)

`git bisect` says:

ed6bdb3c51cd4b8c94f9839806321703e7aa9765 is the first bad commit
commit ed6bdb3c51cd4b8c94f9839806321703e7aa9765
Author: Alexandre Julliard 
Date:   Mon Feb 5 17:03:48 2018 +0100

krnl386: Remove support for DPMI real-mode calls.

Signed-off-by: Alexandre Julliard

$ sha256sum myst_win3x_win.7z
562d3a3d829648352b94ab0d828290f17f90406d34218ebe3ffd0ac012cb12e3

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Alex Henrie  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://gamesnostalgia.com/
                   |                            |download/myst/3403
           Keywords|                            |download, regression

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Alex Henrie  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
    Regression SHA1|                            |ed6bdb3c51cd4b8c94f98398063
                   |                            |21703e7aa9765

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Fabian Maurer  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
           Keywords|                            |win16

--- Comment #1 from Fabian Maurer  ---
This seems to be a program using a DOS extender, I don't think that is
something that wine wants to support anymore.
In the main directly there is a Myst.bat, that works on windows. But under wine
dosbox seems to have issues.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #2 from Alexandre Julliard  ---
It seems to work fine here using the bundled DOSBox.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #3 from Alex Henrie  ---
`wine Myst.bat` works for me too. But is there no way to get MYST.EXE to work
in Wine without DOSBox?

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Zeb Figura  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #4 from Zeb Figura  ---
Wait, so is this an NE program that uses DPMI?

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #5 from Alex Henrie  ---
$ file MYST.EXE
MYST.EXE: MS-DOS executable, NE for MS Windows 3.x (3.10) (EXE)

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #6 from Fabian Maurer  ---
(In reply to Alexandre Julliard from comment #2)
> It seems to work fine here using the bundled DOSBox.

Weird, I don't get any graphical output. It plays some sound after a while, so
I assume it's running, but completely black.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |WONTFIX

--- Comment #7 from Alexandre Julliard  ---
(In reply to Alex Henrie from comment #3)
> `wine Myst.bat` works for me too. But is there no way to get MYST.EXE to
> work in Wine without DOSBox?

There might be, but I don't think we want to invest more effort in DOS support
at this point.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #8 from Alex Henrie  ---
(In reply to Alexandre Julliard from comment #7)
> (In reply to Alex Henrie from comment #3)
> > `wine Myst.bat` works for me too. But is there no way to get MYST.EXE to
> > work in Wine without DOSBox?
>
> There might be, but I don't think we want to invest more effort in DOS
> support at this point.

You've marked this bug as WONTFIX. Maybe it's obvious to everyone else, but
since it's not obvious to me, I'll ask: Why does Wine now have a policy that
forbids supporting NE programs that use DPMI?

Myst was originally distributed on CD without conveniently bundled copies of
Windows 3.1 and DOSBox. Unless users find this bug report, they will have no
idea why the CD doesn't work.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #9 from Alexandre Julliard  ---
(In reply to Alex Henrie from comment #8)
> (In reply to Alexandre Julliard from comment #7)
> > (In reply to Alex Henrie from comment #3)
> > > `wine Myst.bat` works for me too. But is there no way to get MYST.EXE to
> > > work in Wine without DOSBox?
> >
> > There might be, but I don't think we want to invest more effort in DOS
> > support at this point.
>
> You've marked this bug as WONTFIX. Maybe it's obvious to everyone else, but
> since it's not obvious to me, I'll ask: Why does Wine now have a policy that
> forbids supporting NE programs that use DPMI?

If the app is using DPMI, sooner or later it's going to want to call to vm86
mode, which is no longer supported on any recent system. It doesn't work on
Windows either, which is why the app has to be shipped with DOSBox.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #10 from Fabian Maurer  ---
> It doesn't work on Windows either, which is why the app has to be shipped with DOSBox.

I think that's because 64bit windows doesn't support 16Bit code at all.

FWIW the game seems to run without problems in my XP VM. Not sure if that means
XP supports vm86 or that the program doesn't use it.

It also works on otvdm, but that uses emulation.

[Bug 56225] Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #11 from Alexandre Julliard  ---
It is using DPMI calls to invoke DOS interrupts in real mode. It may be
possible to emulate this, but I don't think it's worth the trouble.

FWIW the contents of the cd directory, which I assume is the original CD,
install and work fine here.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Alex Henrie  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Myst deadlocks on start     |16-bit Myst deadlocks on
                   |since Wine 3.2              |start since Wine 3.2

--- Comment #12 from Alex Henrie  ---
There were several versions of Myst. Later CDs included a 32-bit version that
can be made to work on Wine, but the earlier CDs only included the 16-bit
version, which used to work on Wine and doesn't anymore. One of the early
16-bit-only CDs is available at
https://archive.org/download/myst_20200711/MYST.ISO

I've updated the bug title to clarify that this bug only happens on the 16-bit
versions.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #13 from Alex Henrie  ---
Created attachment 75961
  --> https://bugs.winehq.org/attachment.cgi?id=75961
[PATCH] krnl386: Restore simulated real mode interrupts.

This might not actually be that hard to implement. The attached patch is enough
for 16-bit Myst to start, although Myst is very particular about its installer
being run from the D drive. Steps to install and run:

1. Insert the Myst CD

2. Run `sudo sysctl vm.mmap_min_addr=0`

3. Run `wine 'D:\install.exe'` - The installer works but hangs at the end, so
press Ctrl+C when it's done

4. In one terminal, run `Xephyr :1 -ac -screen 800x600x8`

5. In another terminal, run `DISPLAY=:1 openbox`

6. In a third terminal, run `DISPLAY=:1 wine 'C:\MYST\MYST.EXE'`

Unfortunately, Myst deadlocks again when loading the first level. It's not
clear to me whether that's another problem with DPMI or something else
entirely. The first level doesn't load in Wine 3.1 either.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #14 from Fabian Maurer  ---
> Unfortunately, Myst deadlocks again when loading the first level. It's not
> clear to me whether that's another problem with DPMI or something else
> entirely. The first level doesn't load in Wine 3.1 either.

Yeah, the main thread enters a deadloop waiting for I-don't-know-what to
happen.
I tried to debug it, but I can't even get gdb to disassemble the code that
runs. Win16 is a pain.

Not sure it is worth the effort though, you can just install otvdm (e.g.
https://github.com/Winetricks/winetricks/pull/2180) and it runs just fine.
Do you think there is an advantage to have this working natively in wine?

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #15 from Alex Henrie  ---
It would be nice to have 16-bit Myst running in Wine without having to use
Winetricks or other hacks. For one thing, the people who have these old CDs and
want to use them are not necessarily going to be aware that some are 16-bit and
others are 32-bit. And for those of us who are aware of the differences, it
would be interesting to run the different versions of Myst in Wine just to
compare them. Yet another motivation is that Myst is a good example of an NE
program that uses DPMI, so debugging it could help us to debug similar
programs.

If someone is willing to put in the effort to support it, why not let them?

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #16 from Fabian Maurer  ---
> If someone is willing to put in the effort to support it, why not let them?

Heh true, thanks for the writeup. I just always question myself if I don't
waste my time, but I can't let it go for some reason :)

Split of the freeze into bug 56260, more info there.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #17 from Fabian Maurer  ---
Will you be trying to get your patch upstreamed? IMHO that should be
acceptable, although I have to hope Alexandre agrees on that.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #18 from Alex Henrie  ---
I sent it: https://gitlab.winehq.org/wine/wine/-/merge_requests/5052

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #19 from Fabian Maurer  ---
Are you working on an updated version of this? I don't think I understand how
Alexandre wants it implemented.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #20 from Alex Henrie  ---
(In reply to Fabian Maurer from comment #19)
> Are you working on an updated version of this? I don't think I understand
> how Alexandre wants it implemented.

I think Alexandre's comment at
https://gitlab.winehq.org/wine/wine/-/merge_requests/5052#note_60608 primarily
means that he wants Bug 56260 to be fixed first. It may be possible to simplify
the patch further, but since the real-mode contexts and pointers are coming
from Myst, I don't see how it would be possible to not use real-mode contexts
or pointers at all. It seems to me that the best we can do is translate the
real-mode segment pointers to flat pointers to the first megabyte of address
space and run the interrupt handler in protected mode.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #21 from Fabian Maurer  ---
I think it(In reply to Alex Henrie from comment #20)
> I think Alexandre's comment [...]
> primarily means that he wants Bug 56260 to be fixed first. It may be
> possible to simplify the patch further, but since the real-mode contexts and
> pointers are coming from Myst, I don't see how it would be possible to not
> use real-mode contexts or pointers at all.

Yeah me neither, but it sounds like there is another option. At least that's
how I understand it, maybe we should ask for clarification?

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #22 from Fabian Maurer  ---
Are you still working on trying to get this fix upstreamed?

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #23 from Alex Henrie  ---
(In reply to Fabian Maurer from comment #22)
> Are you still working on trying to get this fix upstreamed?

I just noticed that GitLab reported a conflict on
https://gitlab.winehq.org/wine/wine/-/merge_requests/5238 that it couldn't
resolve automatically, so I rebased it just now. I'm still holding out hope
that both that MR and https://gitlab.winehq.org/wine/wine/-/merge_requests/5052
will be accepted.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #24 from Fabian Maurer  ---
I see, thanks for the update. I have to say having only that comment on the
matter, I'm not too hopeful. I'd like to help get this along, but I'm afraid I
don't know what I could do.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

--- Comment #25 from Austin English  ---
Closing.

[Bug 56225] 16-bit Myst deadlocks on start since Wine 3.2

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56225

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #26 from Austin English  ---
Actually closing.