[Bug 54094] New: kernel32:process - test_services_exe() sometimes fails due to an invalid buffer size in Windows

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=54094

Bug ID: 54094
           Summary: kernel32:process - test_services_exe() sometimes fails
                    due to an invalid buffer size in Windows
           Product: Wine
           Version: unspecified
          Hardware: x86-64
                OS: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: kernel32
          Assignee: [e-mail]
          Reporter: [e-mail]

test_services_exe() sometimes fails due to an invalid buffer size in Windows:

process.c:4962: Test failed: got 0xc0000004

See https://test.winehq.org/data/patterns.html#kernel32:process

Where 0xc0000004 == STATUS_INFO_LENGTH_MISMATCH

This causes a little bit over one failure per month.

I think the failure simply happens when a process is started between the first
call to NtQuerySystemInformation(), which returns the buffer size to hold the
*array* of SYSTEM_PROCESS_INFORMATION structures, and the time we call it again
with that size. Obviously we don't start new processes ourselves so it must be
a scheduled task or some equivalent.

We should probably loop a couple of times to defend against that.

[Bug 54094] kernel32:process - test_services_exe() sometimes fails due to an invalid buffer size in Windows

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=54094

François Gouget  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |source, testcase

--- Comment #1 from François Gouget  ---
test_services_exe() sometimes gets an extra failure:

process.c:4981: Test failed: got 0xc0000004
process.c:4999: Test failed: got services.exe SessionId 1

There are only 2 instances of this extra failure over a 6 months period:
* 2022-08-28 on w1064-adm (21H2, 64-bit)
* 2022-11-28 on w10pro64-ru (21H1, 64-bit)

Both times they happened together with the 0xc0000004 failure so I think they
are tied.

I'm hoping we get a bad session id because we use uninitialized data from the
buffer set up by the failed NtQuerySystemInformation() call. If so fixing the
former failure will fix the latter.

[Bug 54094] kernel32:process - test_services_exe() sometimes fails due to an invalid buffer size in Windows

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=54094

François Gouget  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
      Fixed by SHA1|                            |508b793ec915bae900edb42e10d
                   |                            |7bf8d659cd32a
             Status|NEW                         |RESOLVED

--- Comment #2 from François Gouget  ---
This got fixed by the commit below:

commit 508b793ec915bae900edb42e10d7bf8d659cd32a
Author: François Gouget 
Date:   Thu Dec 8 03:39:39 2022 +0100

kernel32/tests: Avoid failures when processes start during the process
test.

On Windows processes sometimes start during test_services_exe() so that
    the size returned by the first NtQuerySystemInformation() is no longer
    sufficient for the second call.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54094

[Bug 54094] kernel32:process - test_services_exe() sometimes fails due to an invalid buffer size in Windows

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=54094

Alexandre Julliard  changed:

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

--- Comment #3 from Alexandre Julliard  ---
Closing bugs fixed in 8.0-rc2.