[Bug 45845] New: Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Bug ID: 45845
           Summary: Microsoft Office 2010: installer crashes at the
                    beginning of the installation process
           Product: Wine
           Version: 3.16
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: msi
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 62317
  --> https://bugs.winehq.org/attachment.cgi?id=62317
Backtrace and CLI-output that were received during installation process

Debian 9.5 Stable (64 bit)
Wine-Development v3.16
Clean Windows 7 prefix (32 bit)

Starting with version 3.15, the MS-Office 2010 crashes during installation (at
the beginning, but not immediately after starting). On Windows XP prefix the
same situation.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Rosanne DiMesio  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |45208
     Ever confirmed|0                           |1
    Regression SHA1|                            |baea371c3d4153bc4a718c07f46
                   |                            |971d4d39aa9f3
             Status|UNCONFIRMED                 |NEW
                 CC|                            |[e-mail]

--- Comment #1 from Rosanne DiMesio  ---
Confirming the early crash, though I suspect this is going to be resolved as a
duplicate of bug 45208.

The regression test showed:

baea371c3d4153bc4a718c07f46971d4d39aa9f3 is the first bad commit
commit baea371c3d4153bc4a718c07f46971d4d39aa9f3
Author: Zebediah Figura 
Date:   Fri Sep 7 20:17:22 2018 -0500

msi: Add exception handling around all custom action RPC calls.

Signed-off-by: Zebediah Figura 
    Signed-off-by: Hans Leidekker 
    Signed-off-by: Alexandre Julliard

:040000 040000 aac7b8f9a1511d39e31a9f6bfdacefb458d04a15
8100fd9e2780398bc325c94d2ca992ec77779570 M      dlls

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #2 from Dmitry Timoshkov  ---
One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
it makes MSI APIs return wrong value on exception. Let's look for instance
at MsiEvaluateConditionW implementation: it's supposed to return MSICONDITION,
however in the case of an exception it returns the exception code instead of
MSICONDITION_NONE.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Dmitry Timoshkov  changed:

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

--- Comment #3 from Dmitry Timoshkov  ---
*** Bug 45850 has been marked as a duplicate of this bug. ***

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #4 from Dmitry Timoshkov  ---
(In reply to Dmitry Timoshkov from comment #2)
> One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
> it makes MSI APIs return wrong value on exception. Let's look for instance
> at MsiEvaluateConditionW implementation: it's supposed to return
> MSICONDITION,
> however in the case of an exception it returns the exception code instead of
> MSICONDITION_NONE.

MsiSetMode() is even more broken regarding the returned value.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #5 from Zebediah Figura  ---
(In reply to Dmitry Timoshkov from comment #2)
> One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
> it makes MSI APIs return wrong value on exception. Let's look for instance
> at MsiEvaluateConditionW implementation: it's supposed to return
> MSICONDITION,
> however in the case of an exception it returns the exception code instead of
> MSICONDITION_NONE.

Eh, well, that's not always a guarantee. For example, COM proxies always return
an HRESULT-style error code even if the function doesn't return an HRESULT.

I'd be very curious to see how either of those functions are throwing an
exception, however.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #6 from Dmitry Timoshkov  ---
(In reply to Dmitry Timoshkov from comment #4)
> > One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
> > it makes MSI APIs return wrong value on exception. Let's look for instance
> > at MsiEvaluateConditionW implementation: it's supposed to return
> > MSICONDITION,
> > however in the case of an exception it returns the exception code instead of
> > MSICONDITION_NONE.
>
> MsiSetMode() is even more broken regarding the returned value.

Add to the list MsiGetLanguage(), MsiProcessMessage(),

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #7 from Dmitry Timoshkov  ---
(In reply to Zebediah Figura from comment #5)
> > One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
> > it makes MSI APIs return wrong value on exception. Let's look for instance
> > at MsiEvaluateConditionW implementation: it's supposed to return
> > MSICONDITION,
> > however in the case of an exception it returns the exception code instead of
> > MSICONDITION_NONE.
>
> Eh, well, that's not always a guarantee. For example, COM proxies always
> return an HRESULT-style error code even if the function doesn't return an
> HRESULT.
>
> I'd be very curious to see how either of those functions are throwing an
> exception, however.

My point is that your patch has changed the return value of many MSI APIs,
and that *might* be the source of the regression.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #8 from Zebediah Figura  ---
(In reply to Dmitry Timoshkov from comment #7)
> My point is that your patch has changed the return value of many MSI APIs,
> and that *might* be the source of the regression.

As far as I'm aware rpcrt4 makes it clear if a proxy call fails. Since I don't
see that in the log, and since I can't really envision a situation where any of
the relevant functions crash, I'm inclined to think that's a red herring.

Anyway, I'll take a look at .NET 4.0.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Austin English  changed:

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

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #9 from Dmitry Timoshkov  ---
I just sent a patch that should fix the regression.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #10 from Dmitry Timoshkov  ---
https://source.winehq.org/patches/data/151322

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Robert Walker  changed:

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

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

[e-mail] changed:

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

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #11 from Rosanne DiMesio  ---
The patch fixes this bug and bug 45850.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Rosanne DiMesio  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
      Fixed by SHA1|                            |09e4240f33425eda3ea05eb2a98
                   |                            |adaec184b7ba4
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #12 from Rosanne DiMesio  ---
The patch was committed; marking fixed.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #13 from Dmitry Timoshkov  ---
(In reply to Dmitry Timoshkov from comment #6)
> > > One of the problems with baea371c3d4153bc4a718c07f46971d4d39aa9f3 is that
> > > it makes MSI APIs return wrong value on exception. Let's look for instance
> > > at MsiEvaluateConditionW implementation: it's supposed to return
> > > MSICONDITION,
> > > however in the case of an exception it returns the exception code instead of
> > > MSICONDITION_NONE.
> >
> > MsiSetMode() is even more broken regarding the returned value.
>
> Add to the list MsiGetLanguage(), MsiProcessMessage().

While this particular regression was not caused by returning wrong value
due an exception, it's still worth IMO to add the tests to the listed above
APIs to check their behaviour.

Zebediah, since that's your patch are you planning to investigate this?

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #14 from Zebediah Figura  ---
(In reply to Dmitry Timoshkov from comment #13)
> While this particular regression was not caused by returning wrong value
> due an exception, it's still worth IMO to add the tests to the listed above
> APIs to check their behaviour.
>
> Zebediah, since that's your patch are you planning to investigate this?

I'm not sure how to force any of those APIs to cause a page fault on Windows.
Do you have any ideas?

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #15 from Dmitry Timoshkov  ---
(In reply to Zebediah Figura from comment #14)
> > While this particular regression was not caused by returning wrong value
> > due an exception, it's still worth IMO to add the tests to the listed above
> > APIs to check their behaviour.
> >
> > Zebediah, since that's your patch are you planning to investigate this?
>
> I'm not sure how to force any of those APIs to cause a page fault on
> Windows. Do you have any ideas?

Setting a break point on an rpcrt4 API that MSI uses for calling to
the remote side seems a workable solution. A starting point would be
to figure out first which one to exploit for this purpose.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

--- Comment #16 from Dmitry Timoshkov  ---
Created attachment 62361
  --> https://bugs.winehq.org/attachment.cgi?id=62361
hack

I played with the attached patch. Unfortunately it looks like any exception
in the thunk leads to the installation termination. Perhaps I'm missing
something, fill free to join the club.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Anastasius Focht  changed:

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

--- Comment #17 from Anastasius Focht  ---
*** Bug 45835 has been marked as a duplicate of this bug. ***

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Alexandre Julliard  changed:

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

--- Comment #18 from Alexandre Julliard  ---
Closing bugs fixed in 3.17.

[Bug 45845] Microsoft Office 2010: installer crashes at the beginning of the installation process

https://bugs.winehq.org/show_bug.cgi?id=45845

Gijs Vermeulen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression