[Bug 16162] New: Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Summary: Settlers 3 crashes with "wine client error: partial
                    write"
           Product: Wine
           Version: 1.1.9
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Settlers 3 causes a wine client error immediately when run. Tested with all
available versions (demo, original, mission pack, QOTA, also with the latest
patch).

$ wine s3demo.exe
wine client error:2d: partial write 64

The demo version causes the same error. The semi-official download locations
did not work for me anymore, but the installer seems to be uploaded here:
http://www.sendspace.com/file/nrunbh

(The installer may complain about disk space but works anyway.)

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #1 from Austin English   2008-11-23 03:14:53 ---
Please attach terminal output.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #2 from Lauri Kenttä   2008-11-23 13:04:17 ---
It doesn't output anything but that, and setting WINEDEBUG=+all or adding own
debugging output to the code (even empty ERR lines) causes a page fault
instead. Also, using a debugger causes some other crashes at some other points
depending on exact usage, usually a stack overflow or a privileged instruction.

I seem to remember that in the past (a year or two ago) the game managed to
start and draw the menu screen once before hanging with timeout in
WaitFor-something.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #3 from Austin English   2008-11-23 13:28:24 ---
(In reply to comment #2)
> It doesn't output anything but that, and setting WINEDEBUG=+all or adding own
> debugging output to the code (even empty ERR lines) causes a page fault
> instead. Also, using a debugger causes some other crashes at some other points
> depending on exact usage, usually a stack overflow or a privileged instruction.
>
> I seem to remember that in the past (a year or two ago) the game managed to
> start and draw the menu screen once before hanging with timeout in
> WaitFor-something.

A regression test would help here:
http://wiki.winehq.org/RegressionTesting

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #4 from Lauri Kenttä   2008-11-23 19:03:33 ---
(In reply to comment #3)
> A regression test would help here:
> http://wiki.winehq.org/RegressionTesting

It's broken in 1.0-rc4, and that's as far as I can test it, because "make"
fails with older commits with this and a few similar errors:
signal_i386.c:613: error: ‘VIF_MASK’ undeclared (first use in this
function)

I tried building some releases between 0.9.10 and 1.0-rc3, and they all caused
the same build error. Any ideas? Versions 0.9.9 and below had some other
problems already in "make depend". Those might fail because of newer build
software, and that's something I won't even try to change.

I wouldn't be so eager to classify this as a regression. It's quite possible
that there has previously been some stubs that have "worked" enough to make the
game start without ever getting to this bug. (I mean something like "fopen
didn't really work so a bug in fclose didn't show up.")

Interestingly, the game is reported to work (AppDB: Bronze) with several of the
wine versions that didn't work for me.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #5 from Austin English   2008-11-23 20:32:24 ---
(In reply to comment #4)
> (In reply to comment #3)
> > A regression test would help here:
> > http://wiki.winehq.org/RegressionTesting
>
> It's broken in 1.0-rc4, and that's as far as I can test it, because "make"
> fails with older commits with this and a few similar errors:
> signal_i386.c:613: error: ‘VIF_MASK’ undeclared (first use in this
> function)
>
> I tried building some releases between 0.9.10 and 1.0-rc3, and they all caused
> the same build error. Any ideas? Versions 0.9.9 and below had some other
> problems already in "make depend". Those might fail because of newer build
> software, and that's something I won't even try to change.

http://wiki.winehq.org/ReverseRegressionTesting

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #6 from Lauri Kenttä   2009-02-15 08:15:26 ---
I tested this in Ubuntu 8.04 with various recent Wine versions, and there it
gets to the main menu (but fps is around 0.1, so it isn't very usable). In Arch
Linux I only get the "partial write". Do you have any ideas about what library
differences could be causing this?

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Colin Pitrat  changed:

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

--- Comment #7 from Colin Pitrat   2009-02-17 14:41:14 ---
I confirm that I have the same problem under archlinux with wine 1.1.15.
> From wine sources, it seems that the "partial write: 64" comes from a call to

writev in (ntdll/server.c):
static unsigned int send_request( const struct __server_request_info *req )

Strange thing is, from strace output, it seems the writev send 1 buffer of
length 64, which means the whole content is written:
writev(3,
[{"\32\0\0\0\230\3763\0\320\3763\0(\0\0\0\1\0\0\0\0@\375\177\30\30\350~\0\0\0\0\0"...,
64}], 1) = 64

Looking at the source code
if ((ret = writev( ntdll_get_thread_data()->request_fd, vec, i+1 )) ==
     req->u.req.request_header.request_size + sizeof(req->u.req)) return
STATUS_SUCCESS;

it would mean that the left hand of the comparison does not correspond to the
size of the buffer. As sizeof(req->u.req) is the 64 passed to writev, it may
comes from req->u.req.request_header.request_size being non-zero.

This is confirmed by the fact that we're in the else part of a
if(!req->u.req.request_header.request_size)

To sum up, it seems that only the header is sent, i.e. the request data_count
is 0 whereas the request_size is not 0. I don't know where this invalid request
could come from.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #8 from Colin Pitrat   2009-02-17 14:53:51 ---
> From winedbg, the call to wine_server_call could be done from the following

stack:
Wine-dbg>bt                              
Backtrace:                               
=>0 0xb80b2422 (0x7ec0364c)              
  1 0x7efc1fa5 NTDLL_wait_for_multiple_objects+0x275() in ntdll (0x7ec0386c)
  2 0x7efc22a8 NtWaitForMultipleObjects+0x68() in ntdll (0x7ec0389c)        
  3 0x7ee5aff2 WaitForMultipleObjectsEx+0x102() in kernel32 (0x7ec039dc)    
  4 0x7ee5b1ec WaitForSingleObject+0x3c() in kernel32 (0x7ec039fc)          
  5 0x007b2bcf in s3 (+0x3b2bcf) (0x7ec03a38)                               
  6 0x7efc5322 call_thread_func+0x42() in ntdll (0x7ec03ad8)                
  7 0x7efc550e start_thread+0x14e() in ntdll (0x7ec043c8)                   
  8 0xb7f24155 start_thread+0xc5() in libpthread.so.0 (0x7ec044b8)          
  9 0xb7ea7a5e __clone+0x5e() in libc.so.6 (0x00000000)

However, I'm not familiar enough with winedbg to check content of the request
object.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #9 from Colin Pitrat   2009-02-18 14:38:36 ---
After deeper analysis, it appears that the stack I got in winedbg is invalid.
The "invalid" request come from 
NTSTATUS WINAPI NtSetEvent( HANDLE handle, PULONG NumberOfThreadsReleased )

It has a data_count == 0, and a request_size that seems to not be initialized.

A call to backtrace / backtrace_symbols_fd give the following stack:
/home/colin/Programmation/wine-1.1.15/loader/../dlls/ntdll/ntdll.dll.so(wine_server_call+0xe5)[0x7efbc735]
/home/colin/Programmation/wine-1.1.15/loader/../dlls/ntdll/ntdll.dll.so(NtSetEvent+0x6b)[0x7efc2afb]
/home/colin/Programmation/wine-1.1.15/loader/../dlls/kernel32/kernel32.dll.so(SetEvent+0x1a)[0x7ee4fa6a]
[0x7b2a77]
[0xeb835b00]

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Karl Smith  changed:

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

--- Comment #10 from Karl Smith   2009-04-24 07:46:13 ---
This happened on my original install. After patching using this file
ftp://ftp.bluebyte.com/updates/s3.exe (I believe you can just drop it into your
settlers directory), everything worked.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #11 from Lauri Kenttä   2009-04-24 11:57:54 ---
(In reply to comment #10)
> After patching using this file ftp://ftp.bluebyte.com/updates/s3.exe
> everything worked.

The Windows XP patch indeed makes the game start. (I probably had that one
installed a few years ago. This time I didn't find it myself, thanks for the
link.)

But this patch doesn't solve everything: Both the Map Editor installer in the
Mission CD and the QOTA installer hang with this partial write error when
accepting the EULA.

Maybe I'll try to hack some kind of custom install script for those.

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Austin English  changed:

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

--- Comment #12 from Austin English   2009-07-21 11:44:28 ---
*** Bug 19333 has been marked as a duplicate of this bug. ***

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

--- Comment #13 from Jeff Zaroyko   2010-06-28 08:28:35 ---
Is this still an issue in Wine-1.2-rc5 or newer?

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Lauri Kenttä  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #14 from Lauri Kenttä   2010-07-05 10:27:08 ---
Both the demo version and the original full version seem to work now, great! :)

I can't test the expansion pack installers now because I don't have the disks
at hand. I'll open a new bug if there's still something wrong with those.
(Hopefully not.)

[Bug 16162] Settlers 3 crashes with "wine client error: partial write"

http://bugs.winehq.org/show_bug.cgi?id=16162

Alexandre Julliard  changed:

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

--- Comment #15 from Alexandre Julliard   2010-07-09 11:56:59 ---
Closing bugs fixed in 1.2-rc7.