[Bug 19781] New: Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Summary: Visual C++ 2005 Express: -Zi option doesn't work;
                    breaks Firefox build's configure script
           Product: Wine
           Version: 1.1.27
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Keywords: download
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Firefox's configure script quite reasonably tries cl -Zi to
see if -Zi works.  Under Wine, sadly, it doesn't.

To reproduce, you have to get past bug 19779 by copying 
C:/Program Files/Microsoft Visual Studio 8/Common7/IDE/mspdb* 
to
C:/windows/system32.

Then try
  cd ~/.wine/drive_c/Program Files/Microsoft Visual Studio 8/Common7/Tools
  cp ~/hello.c .
  wine cmd
  vsvars32.bat
  cl -Zi hello.c 
This causes mspdbsrv.exe -start -spawn to run and persist,
but it seems that cl can't talk to that service properly;
repeating the command
  cl -Zi hello.c 
yields

Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.42 for 80x86
Copyright (C) Microsoft Corporation.  All rights reserved.

hello.c
fixme:ole:NdrCorrelationInitialize (0x33e48c, 0x33e08c, 1024, 0x0): stub
fixme:netapi32:NetWkstaUserGetInfo Level 1 processing is partially implemented
fixme:advapi:LsaOpenPolicy ((null),0x33dc98,0x00000001,0x33dcb4) stub
fixme:advapi:LsaClose (0xcafe) stub
MSPDBSRV: fatal error: Unable to start server: server already exist
fixme:ole:NdrCorrelationInitialize (0x33e48c, 0x33e08c, 1024, 0x0): stub
fixme:netapi32:NetWkstaUserGetInfo Level 1 processing is partially implemented
fixme:advapi:LsaOpenPolicy ((null),0x33dc98,0x00000001,0x33dcb4) stub
fixme:advapi:LsaClose (0xcafe) stub
hello.c : fatal error C1902: Program database manager mismatch; please check
your installation

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #1 from Dmitry Timoshkov   2009-09-21 04:39:05 ---
Something is going wrong with ntlm_auth helper, so far I'm not sure what.

Perhaps Kai could help here.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #2 from Dmitry Timoshkov   2009-09-23 05:14:03 ---
In order to simplify debugging I've added all environment varibales from
vsvars32.bat to System\CurrentControlSet\Control\Session Manager\Environment,
and run 'mspdbsrv.exe -start -spawn ' and 'cl -Zi hello.c' from separate
terminals.

Dan, could you please generate and attach the +relay,+seh,+tid,+secur32,+ntlm
log to see if that shows the same failure (with ntlm_auth) I have?

Just in case that should be
WINEDEBUG=+relay,+seh,+tid,+secur32,+ntlm wine cl -Zi hello.c &>~/cl.log

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Hans Leidekker  changed:

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

--- Comment #3 from Hans Leidekker   2009-09-23 07:48:20 ---
I can reproduce this here. Looks like ntlm_auth fails because it's
not supplied with a password and it can't retrieve any cached
credentials.

I wonder if such a call sequence would succeed on Windows because
the user has already logged on to the OS?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #4 from Dmitry Timoshkov   2009-09-23 09:18:07 ---
(In reply to comment #3)
> I can reproduce this here. Looks like ntlm_auth fails because it's
> not supplied with a password and it can't retrieve any cached
> credentials.

Yep, exactly the same symptoms here.
> I wonder if such a call sequence would succeed on Windows because
> the user has already logged on to the OS?

I don't see anything wrong with the call sequence, the app does
AcquireCredentialsHandleW() followed by InitializeSecurityContextW().
That's a normal sequence to start with.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|-unknown                    |rpc

--- Comment #5 from Dmitry Timoshkov   2009-09-29 05:45:32 ---
This seems to be an rpcrt4.dll issue. It's RPCRT4_ClientAuthorize() who
calls InitializeSecurityContextW() which in turn fails.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #6 from Dmitry Timoshkov   2009-09-30 07:54:41 ---
The app does:
GetUserNameEx(NameSamCompatible, user_name, &len);
RpcBindingSetAuthInfoExW(&binding, user_name, RPC_C_AUTHN_LEVEL_PKT_PRIVACY,
                         RPC_C_AUTHN_WINNT, NULL, RPC_C_AUTHN_DEFAULT, &qos);
then an RPC call happens which leads to InitializeSecurityContextW() which
fails.

The problem seems to be caused by the fact that the app doesn't pass any
authentication info to RpcBindingSetAuthInfoExW, although MSDN states that

"When using the RPC_C_AUTHN_WINNT authentication service AuthIdentity should
be a pointer to a SEC_WINNT_AUTH_IDENTITY structure" and "Specify a null value
to use the security login context for the current address space."

The app passes NULL as AuthIdentity handle, and RpcBindingSetAuthInfoExW
tries to get an existing authentication credentials, and it appears that
secur32/ntlm_auth can't cope with that.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Dmitry Timoshkov  changed:

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

--- Comment #7 from Dmitry Timoshkov   2009-10-05 00:24:24 ---
*** Bug 20241 has been marked as a duplicate of this bug. ***

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #8 from Hans Leidekker   2009-10-06 03:34:19 ---
I've submitted a test showing that InitializeSecurityContext
returns SEC_I_CONTINUE_NEEDED after supplying null authentication
data to AcquireCredentialsHandle.

The problem is that on Wine there's currently no concept of a
"security login context for the current address space", as msdn
calls it. This is what will be used when you pass null authentication
data to AcquireCredentialsHandle.

You can work around this bug by caching credentials, e.g. with code
like this:

CREDENTIALA cred;
    static WCHAR pwd[] = {'p','w','d'};
    static char user[] = {'h','o','s','t','\','u','s','e','r',0};

memset(&cred, 0, sizeof(cred));
    cred.Type               = CRED_TYPE_DOMAIN_PASSWORD;
    cred.TargetName         = user;
    cred.CredentialBlobSize = sizeof(pwd);
    cred.CredentialBlob     = (LPBYTE)pwd;
    cred.Persist            = CRED_PERSIST_SESSION;
    cred.UserName           = user;

CredWriteA(&cred, 0);

Substituting static data appropriately, of course. Now
InitializeSecurityContext will return SEC_I_CONTINUE_NEEDED
and we run into another crash:

001a:Ret  secur32.EncryptMessage() retval=80090321 ret=7eb3f895
001a:err:rpc:RPCRT4_SecurePacket EncryptMessage failed with 0x80090321
001a:Call ntdll.RtlFreeHeap(00110000,00000000,001699b8) ret=7eb408ce
001a:Ret  ntdll.RtlFreeHeap() retval=00000001 ret=7eb408ce
001a:Call ntdll.RtlFreeHeap(00110000,00000000,00169b10) ret=7eb3f356
001a:Ret  ntdll.RtlFreeHeap() retval=00000001 ret=7eb3f356
001a:Call KERNEL32.RaiseException(00000721,00000000,00000000,00000000)
ret=7eb4d405

80090321 == SEC_E_BUFFER_TOO_SMALL. Is rpcrt4 calling EncryptMessage
too early, i.e. before completing the ntlm handshake?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Alan Jones  changed:

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

--- Comment #9 from Alan Jones   2009-10-08 23:09:37 ---
As would be expected this also occurs with VS2008.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #10 from Kai Blin   2009-10-14 01:13:15 ---
I'll try to get this setup on my system, but while I'm looking into that, could
anybody who already has the setup get me a +ntlm,+secur32 trace?

One other thing.. ntlm_auth doesn't do authentication standalone. Is vc
expecting to have a domain environment around? If it's trying to do NTLM auth
locally, we're in trouble, as we can't do that. We could try and fake something
there, but I haven't seen any calls to AcceptSecurityContext, so I'm not sure
if that's being called at all.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #11 from Dmitry Timoshkov   2009-10-14 02:41:03 ---
(In reply to comment #10)
> I'll try to get this setup on my system, but while I'm looking into that, could
> anybody who already has the setup get me a +ntlm,+secur32 trace?

The test Hans added is supposed to replicate the problem. Does the test
pass for you?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #12 from Dmitry Timoshkov   2009-10-14 02:48:35 ---
Here is the test:
http://source.winehq.org/git/wine.git/?a=commitdiff;h=8bb68933ea37766c3fd4ed...

And yes, the app tries to get an authentication with the local service via rpc.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #13 from Kai Blin   2009-10-14 06:15:52 ---
(In reply to comment #11)
> (In reply to comment #10)
> > I'll try to get this setup on my system, but while I'm looking into that, could
> > anybody who already has the setup get me a +ntlm,+secur32 trace?
>
> The test Hans added is supposed to replicate the problem. Does the test
> pass for you?

Nope. Let me check if this works when I'm authenticated via winbind and I can
get a winbind credential from the cache.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Kai Blin  changed:

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

--- Comment #14 from Kai Blin   2009-10-14 06:36:00 ---
(In reply to comment #11)
> The test Hans added is supposed to replicate the problem. Does the test
> pass for you?

Right, I missed that. Sorry. Nope, as expected it doesn't pass for me either.
This is Rob's cached credentials code, I never got the Wine-side credential
cache to work myself. :) I know of one user who uses the winbind cached
credentials, but the wine cred cache seems to be pretty un-tested (and broken,
no big surprise).

(In reply to comment #12)
> And yes, the app tries to get an authentication with the local service
> via rpc.

I think we could cheat our way through this here. I haven't looked into this
stuff for a while, but all we need to come up with for the local case is a
username/password combination ntlm_auth will accept for the "server" side.
You can easily bypass the usual check ntlm_auth does for the server side by
running ntlm_auth with --username=user --password=pass --domain=dom

So assuming ntlm_GetCachedCredential() returned user WINE\bob and password
secret if there's no real cached credentials... we'll handsomely break
winbind-based cached credentials if we need to authenticate against a remote
server.

I'll have to give this some more thought. Sorry I can't provide you with a
ready solution here.

A long-term fix would be to get Samba winbindd to also be able to authenticate
local users (like the local SAM on a windows machine) and store that user's
creds in the winbind cache at login time, just like it does for domain logins.
Then Wine could always use the winbind credentials cache.
Unfortunately, winbindd can't do that quite yet, probably a couple of man-weeks
worth of work on Wine and Samba if you know where to look.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #15 from Kai Blin   2009-10-14 06:36:31 ---
(In reply to comment #13)
> (In reply to comment #11)
> > (In reply to comment #10)
> > > I'll try to get this setup on my system, but while I'm looking into that, could
> > > anybody who already has the setup get me a +ntlm,+secur32 trace?
> >
> > The test Hans added is supposed to replicate the problem. Does the test
> > pass for you?
>
> Nope. Let me check if this works when I'm authenticated via winbind and I can
> get a winbind credential from the cache.

Scratch that, didn't think about that one enough.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #16 from Kai Blin   2009-10-14 12:47:56 ---
Created an attachment (id=24129)
 --> (http://bugs.winehq.org/attachment.cgi?id=24129)
Use empty passwird if cached credentials fail

Work-around for the failing cached credentials. This sends an empty password to
make ntlm_auth happy. With this workaround, we should be able to get to the
point where the server side of the RPC fails.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #17 from Dmitry Timoshkov   2009-10-15 01:09:44 ---
(In reply to comment #16)
> Work-around for the failing cached credentials. This sends an empty password to
> make ntlm_auth happy. With this workaround, we should be able to get to the
> point where the server side of the RPC fails.

Yes, this patch helps. Now the compiler proceeds to the same point as
in the Comment #8 for Hans.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #18 from Kai Blin   2009-10-15 01:36:57 ---
Can I get a +ntlm trace of that, please?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #19 from Hans Leidekker   2009-10-15 03:02:35 ---
Created an attachment (id=24143)
 --> (http://bugs.winehq.org/attachment.cgi?id=24143)
+ntlm,+secur32,+rpc trace with patch applied

Here's a +ntlm,+secur32,+rpc log. It's not exactly the same failure as
with my workaround, EncryptMessage fails here with SEC_E_BUFFER_TOO_SMALL.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #20 from Dan Kegel   2009-10-21 23:22:43 ---
I'm jonesin' for this; being able to builds the memcheck test
suite under wine with /Zi, and then run it under valgrind+wine
would be quite the hat trick.  What's the current status?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #21 from Kai Blin   2009-10-22 00:52:04 ---
I've got this problem reproduced. The issue seems to be that the RPC code never
finishes the NTLM handshake. I've failed to find the correct place to send the
initial request to the server, the code seems to assume
InitializeSecurityContext takes care of that. While trying to wrap my head
around the RPC code, university caught up with me again.

I know that I demoed Wine doing NTLM crypto with Outlook in 2006. Did the RPC
code significantly change since then? I don't understand how this code would
have done the right thing in 2006.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #22 from Hans Leidekker   2009-10-22 02:17:49 ---
Perhaps you demoed outlook in rpc over https mode? In which case
ntlm authentication works fine. I agree that secure rpc's likely
have not worked before. At least I can't find the code to build all
of the corresponding packets.

I found this old perl implementation, perhaps it's useful:
http://cpansearch.perl.org/src/UMVUE/DCE-RPC-0.11/lib/DCE/RPC.pm

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #23 from Dan Kegel   2009-10-26 19:47:08 ---
Hi Kai,
thanks for looking at this.  Hope university releases its grip enough
to let you wrap your head around it soon!  What's the prognosis?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #24 from Kai Blin   2009-10-29 03:21:41 ---
(In reply to comment #23)
> thanks for looking at this.  Hope university releases its grip enough
> to let you wrap your head around it soon!  What's the prognosis?

Well, _if_ I could find my way around the horrible mess that is our RPC code, I
might be able to do something. I keep wasting the few hours I have per week at
poking around the rpc code.

Basically, RPCRT4_ClientAuthorize needs to send the buffer returned from
InitializeSecurityContext to the server, then receive the reply from the server
and loop on doing that until the result from InitializeSecurityContext isn't
SEC_I_CONTINUE anymore. However, so far I haven't managed to find out how to do
the "send to server" and "receive from server" parts are supposed to work.

I'm stuck at that point.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #25 from Hans Leidekker   2009-11-04 02:44:46 ---
Created an attachment (id=24545)
 --> (http://bugs.winehq.org/attachment.cgi?id=24545)
rpcrt4: Add secure rpc test.

This patch adds a test that reproduces the crash seen in this bug.

It applies on top of this patch:
http://www.winehq.org/pipermail/wine-patches/2009-November/080899.html

My plan was to run the same test over ncacn_ip_tcp on windows and
spy on it with wireshark, but alas, I get an unknown exception 0x0000063d
when I do that. Same thing with ncacn_np, only ncalrpc seems to allow
secure mode.

A possible explanation is that I'm supplying default credentials, which
are only valid on the machine itself.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #26 from Hans Leidekker   2009-11-04 05:04:04 ---
I was wrong, it appears that authentication services need to be registered
by the rpc server and ncalrpc simply has RPC_C_AUTHN_WINNT registered by
default. So, after a call to RpcServerRegisterAuthInfo, my test appears
to do secure rpcs over ncacn_ip_tcp.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #27 from Hans Leidekker   2009-11-05 07:44:57 ---
It's not as bad as I thought, we do support secure rpcs when explicit
credentials are supplied in the call to RpcBindingSetAuthInfoEx.

This means we're back to the problem of handling default credentials.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #28 from Kai Blin   2009-11-07 04:59:42 ---
Actually the problem is that the rpc server side in Wine does not handle
authentication whatsoever. It never calls out to SSPI and never attaches a
challenge packet to the BIND_ACK reply, so the rpc client code never finishes
the handshake.

So technically this is an rpc bug. :)

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Rob Shearman  changed:

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

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #29 from Rob Shearman   2009-11-07 07:23:06 ---
While the app here may benefit from work to add support for authentication on
the server side, I suspect that authentication over ncalrpc is special cased to
be a no-op in the native DLL. This is supported by Hans' findings that
RpcServerRegisterAuthInfo does not need to be called on the server side for the
client to be able to succeed whilst specifying NTLM authentication to be
applied. More tests need to be done to see whether the identity passed into
RpcBindingSetAuthInfo is seen on the server side (RpcBindingInqAuthClient) for
ncalrpc & ncacn_ip_tcp.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Vincent Cappe  changed:

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

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Hans Leidekker  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #24545|0                           |1
        is obsolete|                            |

--- Comment #30 from Hans Leidekker   2009-11-17 04:08:59 ---
Created an attachment (id=24799)
 --> (http://bugs.winehq.org/attachment.cgi?id=24799)
rpcrt4: Add secure rpc tests.
> applied. More tests need to be done to see whether the identity passed into
> RpcBindingSetAuthInfo is seen on the server side (RpcBindingInqAuthClient) for
> ncalrpc & ncacn_ip_tcp.

These updated tests succeed here on Windows and crash in the same way
cl.exe does on Wine. RpcBindingInqAuthClient does not return a handle to
the client identity in the Privs parameter in case of ncalrpc, if that's
what you mean. Instead it returns the principal as a Unicode string, as
documented on MSDN.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #31 from Rob Shearman   2009-11-29 04:52:50 ---
Created an attachment (id=25000)
 --> (http://bugs.winehq.org/attachment.cgi?id=25000)
Patch series add support for RPC server authentication

With this patch series applied:
- ncacn_ip_tcp attempts to authenticate using SSPI, but fails for me because
ntlm_auth can't connect to winbindd (and crashes in DecryptMessage due to a
secur32 bug).
- ncalrpc doesn't use SSPI for authentication, instead relying on the security
built into the LRPC transport.

RpcBindingInqAuthClientA/W still need to be fixed to report the connection as
authenticated when using ncalrpc but I'm hoping this is enough to allow the
application to work for the moment.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #32 from Dan Kegel   2009-11-29 18:53:11 ---
Close, but not quite there yet.  Doing
  rm -rf ~/.wine
  sh winetricks vc2005trial
  cd ~/.wine/drive_c/Program Files/Microsoft Visual Studio 8/Common7/Tools
  cp ~/hello.c .
  wine cmd
  vsvars32.bat
  cl -Zi hello.c 
yields
fixme:ole:NdrCorrelationInitialize (0x33e330, 0x33df30, 1024, 0x0): stub
fixme:ole:NdrCorrelationInitialize (0x33e330, 0x33df30, 1024, 0x0): stub
fixme:rpc:NdrStubCall2 new correlation description not implemented
fixme:rpc:calc_arg_size Unhandled type 30
err:rpc:I_RpcReceive we got fault packet with status 0x3e6
fixme:ole:NdrCorrelationInitialize (0x33e330, 0x33df30, 1024, 0x0): stub
fixme:rpc:NdrStubCall2 new correlation description not implemented
fixme:rpc:calc_arg_size Unhandled type 30
fixme:ole:NdrCorrelationFree (0x33e330): stub
hello.c : fatal error C1902: Program database manager mismatch; please check
your installation

which seems a less dire set of errors than in the original post,
but still the app didn't work.  (I'm pretty sure the same problem
affects both vc2005 trial and vc2005 express.)

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #33 from Rob Shearman   2009-12-12 13:16:48 ---
Created an attachment (id=25184)
 --> (http://bugs.winehq.org/attachment.cgi?id=25184)
Patch series to allow MS PDB server to work

Patch series does:
1. Adds support for context handles in stubless server code. This fixes the
following messages:
  fixme:rpc:calc_arg_size Unhandled type 30
  err:rpc:I_RpcReceive we got fault packet with status 0x3e6
2. Add support for [out]-only non-conformant strings. This fixes the following
message after the above is fixed:
  fixme:rpc:calc_arg_size Unhandled type 25

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Ian McLean  changed:

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

--- Comment #34 from Ian McLean   2009-12-12 17:00:41 ---
I've confirmed that Rob's two latest patches enable PDB generation with:
    MSPDB71.DLL version 7.10.6030
    MSPDB80.DLL version 8.0.50727.762
    MSPDB80.DLL version 9.0.30729.1

Built with:
    CL.EXE version 13.10.6030
    CL.EXE version 14.00.50727.762
    CL.EXE version 15.00.30729.01

AKA:
    Microsoft Visual Studio .NET 2003
    Microsoft Visual Studio 2005
    Microsoft Visual Studio 2008

Non-trivial test; over 500MB of PDBs; many applications and DLLs.

Thank you!  This closes our cross-compile loop for development builds.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #35 from Dan Kegel   2009-12-12 17:28:17 ---
w00t!   Looking forward to seeing these on wine-patches...

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #36 from Dan Kegel   2009-12-16 02:24:01 ---
I see some patches are committed.  Is this fixed in git, then?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Rob Shearman  changed:

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

--- Comment #37 from Rob Shearman   2009-12-27 07:35:32 ---
All of the patches I attached are now committed so marking this as fixed.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Dan Kegel  changed:

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

--- Comment #38 from Dan Kegel   2010-01-04 16:25:47 ---
Doesn't seem fixed here today.  I did 'winetricks msvc2005trial', then 
  wine cmd
  vsvars32.bat
  cl -Zi hello.c 
or equivalently
  cd ~/.wine/drive_c/Program Files/Microsoft Visual Studio 8/Common7/IDE
  wine cl -Zi hello.c 
but it failed with
  fixme:rpc:RpcBindingSetAuthInfoExW unsupported AuthnSvc 10
  hello.c : fatal error C1902: Program database manager mismatch; please check
your installation

:-(
This was with this morning's git.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #39 from Dan Kegel   2010-01-04 16:36:50 ---
For completeness, here's the +rpc log:

trace:rpc:RpcStringBindingComposeW
((null),L"ncalrpc",(null),L"mspdb_8.00.50727.42_rtl_32_0000000000000000",(null),0x32e628)
trace:rpc:RpcBindingFromStringBindingW
(L"ncalrpc:[mspdb_8.00.50727.42_rtl_32_0000000000000000]",0x103296bc)
trace:rpc:RpcStringBindingParseW
(L"ncalrpc:[mspdb_8.00.50727.42_rtl_32_0000000000000000]",0x32e5dc,0x32e5d8,0x32e5d4,0x32e5d0,0x32e5cc)
trace:rpc:RPCRT4_CreateBindingW binding: 0x150648
trace:rpc:RPCRT4_SetBindingObject (*RpcBinding == ^0x150648, UUID ==
{00000000-0000-0000-0000-000000000000})
trace:rpc:RPCRT4_CompleteBindingW (RpcBinding == ^0x150648, NetworkAddr == L"",
EndPoint == L"mspdb_8.00.50727.42_rtl_32_0000000000000000", NetworkOptions ==
(null))
trace:rpc:RPCRT4_GetAssociation new assoc 0x150c78
trace:rpc:RpcBindingSetAuthInfoExW 0x150648 L"ubuntu\dank" 6 10 (nil)
4294967295 0x32d5e8
trace:rpc:RpcBindingSetAuthInfoExW SecurityQos { Version=1, Capabilities=0x1,
IdentityTracking=0, ImpersonationLevel=3}
fixme:rpc:RpcBindingSetAuthInfoExW unsupported AuthnSvc 10
hello.c : fatal error C1902: Program database manager mismatch; please check
your installation

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #40 from Dan Kegel   2010-01-04 16:48:45 ---
Happens with both Visual C++ 2005 Trial (which takes ages to install)
and Visual C++ 2005 Express (which is quite quick to install with
  winetricks -q vc2005express ).

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Dan Kegel  changed:

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

--- Comment #41 from Dan Kegel   2010-01-05 05:16:14 ---
Rob wrote:

--- snip ---
AuthnSvc 10 is:
#define RPC_C_AUTHN_WINNT 10

Either ntlm_auth is missing (or outdated) on this machine or there has
been a regression in secur32. A +ntlm,+secur32 log would probably help
diagnose further.
--- snip ---

I'm not at the machine that had the problem, but I am at another
good system, and it's not missing ntlm_auth, so I gave it a whirl,
and cl -Zi seems to work here.

Maybe we need a winediag log message to catch this case.  (In the
past, having a missing ntlm_auth module was a no-op, so if there's
a warning, people's eyes are trained to ignore it...)

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #42 from Kai Blin   2010-01-05 05:29:02 ---
> Maybe we need a winediag log message to catch this case.  (In the
> past, having a missing ntlm_auth module was a no-op, so if there's
> a warning, people's eyes are trained to ignore it...)

There should be a very obvious ERR output if ntlm_auth is missing, along with
suggestions on what package to install to fix said problem. If people are
ignoring that error and complain about authentication errors, I don't know how
to fix _that_.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #43 from Dan Kegel   2010-01-05 05:37:30 ---
It's obvious, but it's usually totally safe to ignore, so I do ignore it.

Can you generate that warning only in situations where it really matters?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #44 from Kai Blin   2010-01-05 06:37:12 ---
(In reply to comment #43)
> It's obvious, but it's usually totally safe to ignore, so I do ignore it.
>
> Can you generate that warning only in situations where it really matters?

Currently it's generated when secur32.dll is loaded and fails to initialize the
NTLM provider. That seemed like a sane spot for me.

It might make sense to add another error in the rpcrt at the point where it
expects to use the RPC_C_AUTHN_WINNT AuthSvc and that's not provided. However,
that'll only help for apps that get to the secur32 code via the rpcrt4 code.

Apps that explicitly use secur32 show different patterns of selecting the
provider they want to use, and on secur32.dll load is the only sane spot to
complain about ntlm_auth missing.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #45 from Dan Kegel   2010-01-05 10:23:56 ---
We could keep the warning you put in, and also add more 
urgent winediag warnings at the later points of use that
popular apps hit, maybe?

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

--- Comment #46 from Dan Kegel   2010-01-05 12:18:31 ---
'apt-get install winbind' on the affected machine does indeed solve the
problem.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

http://bugs.winehq.org/show_bug.cgi?id=19781

Jeff Zaroyko  changed:

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

--- Comment #47 from Jeff Zaroyko   2010-01-09 04:46:38 ---
Closing bugs fixed in 1.1.36.

[Bug 19781] Visual C++ 2005 Express: -Zi option doesn't work; breaks Firefox build's configure script

https://bugs.winehq.org/show_bug.cgi?id=19781

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
      Fixed by SHA1|                            |d412bcc3aef1ccec410b82fac74
                   |                            |e4e55897d9438
                 CC|                            |[e-mail]