[Bug 14827] New: Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

Summary: Autocad 2005 : Multiline text edit crashes application
           Product: Wine
           Version: CVS/GIT
          Platform: PC-x86-64
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

If you try to edit a multiline text (MTEXT) or a dimension text on autocad
(tested on 2005), following happens :

1) The editor is (correctly) shown, but edited text is not visible on text area
2) If you use ESC to abort command, you retourns on autocad correctly
3) If you press OK button to accept changes, the whole autocad hangs.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #1 from Lei Zhang   2008-08-11 20:11:47 ---
Can you attach the console output?

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

[e-mail] changed:

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

--- Comment #2 from [e-mail]  2008-08-12 02:54:54 ---
Here the relevant part, just after issuing command :

fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common Unsupported yet non NULL device in
EM_SETTARGETDEVICE
fixme:richedit:RichEditWndProc_common Unsupported yet non NULL device in
EM_SETTARGETDEVICE
fixme:richedit:RichEditWndProc_common Unsupported yet non NULL device in
EM_SETTARGETDEVICE
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common Unsupported yet non NULL device in
EM_SETTARGETDEVICE
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub

The text edit box appears, but no text inside it.
Here what happens after pressing OK button :

fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:IRichEditOle_fnGetObjectCount stub 0xa3160c8
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:IRichEditOle_fnGetObjectCount stub 0xa316090
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub

Then autocad hangs here (no crash, just hang with 100% cpu usage).
Only way to recover it is to kill the process.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #3 from [e-mail]  2008-11-04 17:01:47 ---
Another test, easyly reproductible : enter autocad, issue MTEXT command, select
a text area on screen, you can enter the text. If you press OK button to
accept, the app hangs with 100% cpu; output is :

(ok button pressed)

fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:IRichEditOle_fnGetObjectCount stub 0xa3264a8
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:IRichEditOle_fnGetObjectCount stub 0xa326470
fixme:richedit:ME_StreamOutFlush Invalid returned written size *pcb: 0x1e5
(485) instead of 460
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub
fixme:richedit:RichEditWndProc_common EM_SELECTIONTYPE: stub

(program hangs here).

If you press ESC instead of OK (so canceling the text input) the app behaves
normally.

Latest wine git of today.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

ruelle  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1

--- Comment #4 from ruelle   2009-01-19 17:36:22 ---
*** This bug has been confirmed by popular vote. ***

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

[e-mail] changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|CVS/GIT                     |1.1.13

--- Comment #5 from [e-mail]  2009-01-19 17:49:01 ---
Present since before wine 1.0.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

Dan Kegel  changed:

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

--- Comment #6 from Dan Kegel   2009-01-19 18:40:52 ---
Does 'wget http://kegel.com/wine/winetricks; sh winetricks riched20' work
around the problem?

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #7 from [e-mail]  2009-01-20 07:13:10 ---
(In reply to comment #6)
> Does 'wget http://kegel.com/wine/winetricks; sh winetricks riched20' work
> around the problem?

Nope, but it has a strange side-effect.... when I select OK after editing
multiline text, it starts to load additional autocad modules (!!!) and then
hangs as usual. It's like some bad message get sent to autocad.
Btw, on console the only thing I get after pressing ok is

fixme:resource:GetGuiResources (0xffffffff,0): stub

(of course, with no special debug channel enabled).

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|[e-mail]              |
            Version|1.1.13                      |unspecified

--- Comment #8 from Dmitry Timoshkov   2009-01-20 10:46:14 ---
(In reply to comment #5)
> Present since before wine 1.0.

Then 1.1.13 is wrong number.

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #9 from [e-mail]  2009-04-20 07:30:45 ---
Just to summarize what I've found so far....
1) The problem is NOT due to richedit20. Using window's one hangs the
application about on the same way.
2) Putting some debugging message inside richedit20 code, I could see that the
problem arises only when this message gets printed :

fixme:richedit:ME_StreamOutFlush Invalid returned written size *pcb: 0x1e5
 (485) instead of 460

The numbers can of course change.
The callback function reads some more bytes than asked ones and hangs
somewhere.
The strange stuff is that on windows it works, and the callback function is
(presumably) inside autocad code.....

3) If I make the function ME_StreamOutPrint() a no-op, the app works, of course
the rtf formatting is lost and is instead partially put inside the edited
string. No hangs anymore.

I guess it could be an OLE  problem somewhere, but not sure about. Autocad is
NOT runnable with relay enabled nor it runs under debugger because of copy
protection. I've still not found where exactly the application hangs, and
wether is inside wine or autocad itself.
If somebody could explain me how rtf editor is called by application via OLE I
could maybe find some more....
This is one of the last 3 bugs that keeps autocad away from being be a platinum
app, and the other 2 are non-blocking ones, so it would be nice to have it
solved.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #10 from [e-mail]  2009-04-20 07:42:51 ---
Created an attachment (id=20572)
 --> (http://bugs.winehq.org/attachment.cgi?id=20572)
Some fake FIXME's added to riched's code - Shows where app hangs

For completness I add here a log got with +ole and some debug messages added to
richedit code, just the last part, from editor invoking up to the crash.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #11 from Dan Kegel   2009-04-20 07:44:56 ---
Does 'winetricks dcom98' help?  (It's risky, usually doesn't help these
days, but worth a shot.)
http://appdb.winehq.org/objectManager.php?sClass=version&iId=102&iTe...
shows that autocad 2000 still benefits from dcom98.

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #12 from [e-mail]  2009-04-20 07:48:26 ---
(In reply to comment #11)
> Does 'winetricks dcom98' help?  (It's risky, usually doesn't help these
> days, but worth a shot.)
> http://appdb.winehq.org/objectManager.php?sClass=version&iId=102&iTe...
> shows that autocad 2000 still benefits from dcom98.

Hi Dan :-)
I tested DCOM98 some time ago with autocad2005, it didn't work.... In my small
experience, I've seen that it helps seldom, anyways.
Do you know some way to locate the point in code where app hangs, without relay
nor debugging possible ? That bug's driving me mad.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #13 from Dan Kegel   2009-04-20 07:55:12 ---
You're doing fine, what's missing is more understanding of COM arcana, I'm
afraid.

Try the tip in
http://appdb.winehq.org/objectManager.php?sClass=version&iId=5658
That app also had the "ME_StreamOutFlush Invalid returned written size"
error, and somebody found a workaround, at least for that app:
"Everything works, (!) but only if you install the packaged libraries, into
windows/system/ of wine: COMCAT.DLL -> register with "wine regsvr32
PATH2DLL/COMCAT.DLL" OLEAUT32.DLL -> -"- OLEPRO32.DLL -> -"- STDOLE2.TLB -> no
need to register Furthermore you have to activate these native libraries via
winecfg. "

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #14 from [e-mail]  2009-04-20 08:28:15 ---
(In reply to comment #13)
> You're doing fine, what's missing is more understanding of COM arcana, I'm
> afraid.
>
> Try the tip in
> http://appdb.winehq.org/objectManager.php?sClass=version&iId=5658
> That app also had the "ME_StreamOutFlush Invalid returned written size"
> error, and somebody found a workaround, at least for that app:
> "Everything works, (!) but only if you install the packaged libraries, into
> windows/system/ of wine: COMCAT.DLL -> register with "wine regsvr32
> PATH2DLL/COMCAT.DLL" OLEAUT32.DLL -> -"- OLEPRO32.DLL -> -"- STDOLE2.TLB -> no
> need to register Furthermore you have to activate these native libraries via
> winecfg. "

Thanx for the link.
I looked at it, but the behaviour is different.... This app don't set the *pcb
field when returning from callback, leaving so the original value (0xDEADBEEF,
nice default, indeed, even if 0xDEADC0DE would have been more appropriate :-)
).
In autocad, the app DOES set it, but with a somehow bigger value than what
requested.... So, I guess the callback do indeed read more bytes than
necessary.
The strange stuff is that the callback should be (I think....) inside autocad
code.... Because of that I'd need to know where exactly the app hangs.
Debugging OLE without knowing its internals would be a nightmare, I guess.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #15 from [e-mail]  2009-04-23 17:14:56 ---
Created an attachment (id=20645)
 --> (http://bugs.winehq.org/attachment.cgi?id=20645)
Workaround to MTEXT crash

Suppressing font table stream out inside dlls/riched20/writer.c makes AutoCAD
happy with multiline text.
Strange stuff..... I'll investigate a bit more.
In the meanwhile, it works quite well as a workaround..... here the related
hacky patch.
Beware, it's just a workaround and I don't know if it'll work for you....

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #16 from Austin English   2009-04-23 17:50:10 ---
Please attach a +richedit trace.

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

Dylan Smith  changed:

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

--- Comment #17 from Dylan Smith   2009-04-23 20:55:04 ---
(In reply to comment #9)
> - Putting some debugging message inside richedit20 code, I could see that the
>
> problem arises only when this message gets printed :
>
> fixme:richedit:ME_StreamOutFlush Invalid returned written size *pcb: 0x1e5
>  (485) instead of 460
>
> The numbers can of course change.
> The callback function reads some more bytes than asked ones and hangs
> somewhere.

I don't have AutoCAD 2005 to test, so I can will make a guess as to why the
callback function may read more bytes than the amount written. The callback may
be looking for a NULL byte to determine the amount of bytes written, instead of
looking at the using the cb value.

Try making the following change to add the NULL byte that native richedit
controls include, and see if the fixme message still gets printed for the *pcd
value.

diff --git a/dlls/riched20/writer.c b/dlls/riched20/writer.c
index d1d9990..17134b4 100644
--- a/dlls/riched20/writer.c
+++ b/dlls/riched20/writer.c
@@ -904,7 +904,7 @@ ME_StreamOutRTF(ME_TextEditor *editor, ME_OutStream
*pStream, int nStart, int nC
       break;
     p = ME_FindItemFwd(p, diRunOrParagraphOrEnd);
   }
-  if (!ME_StreamOutPrint(pStream, "}"))
+  if (!ME_StreamOutPrint(pStream, "}\0"))
     return FALSE;
   return TRUE;
 }

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #18 from Dylan Smith   2009-04-23 21:06:03 ---
Sorry, the previous patch didn't end up doing what I was trying to do, since I
needed the length of the string to be specified.  Try the following instead.

diff --git a/dlls/riched20/writer.c b/dlls/riched20/writer.c
index d1d9990..b6c3ed6 100644
--- a/dlls/riched20/writer.c
+++ b/dlls/riched20/writer.c
@@ -904,7 +904,7 @@ ME_StreamOutRTF(ME_TextEditor *editor, ME_OutStream
*pStream
       break;
     p = ME_FindItemFwd(p, diRunOrParagraphOrEnd);
   }
-  if (!ME_StreamOutPrint(pStream, "}"))
+  if (!ME_StreamOutMove(pStream, "}\0", 2))
     return FALSE;
   return TRUE;
 }

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #19 from [e-mail]  2009-04-24 01:46:32 ---
(In reply to comment #17)
> I don't have AutoCAD 2005 to test, so I can will make a guess as to why the
> callback function may read more bytes than the amount written. The callback may
> be looking for a NULL byte to determine the amount of bytes written, instead of
> looking at the using the cb value.
>
> Try making the following change to add the NULL byte that native richedit
> controls include, and see if the fixme message still gets printed for the *pcd
> value.

......

eheheh... that was the first stuff I've thought.... It works, I mean, adding
the NULL byte make app read the correct amount of bytes (so, I guess that one
is the correct behaviour) but it hangs anyways with 100% cpu as before.
So, I guess the real bug is another one, even if the null byte is (IMHO) still
needed.

I've tryed many stuffs, from a complete +rich*** log, +ole*** log, many
embedded print messages in all calls (ole has very few TRACEs, indeed....), but
I couldn't
manage to make it work. Richedit (and OLE) logs just shows that the app hangs
returning from message handling function, but I couldn't find if it's inside
Autocad (I think so...) or inside wine code.
Autocad don't run with +relay, and the usage of a BackTrace() embedded function
 ( seet devel mailing list) didn't help too much.

@Austin : the log (with many printf() messages added) is already here, the
first attachment. As you can see, it shows nothing :-(

Suppressing the \fonttbl output makes autocad work almost perfectly (there are
some other unrelated bugs in richedit code).
Native riched20 didn't help either, as already told here.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #20 from Austin English   2009-04-24 10:45:19 ---
(In reply to comment #19)
> @Austin : the log (with many printf() messages added) is already here, the
> first attachment. As you can see, it shows nothing :-(

Unless you changed every single warn/err/trace to a fixme, you're missing some
output. Running with +richedit it much easier/faster.
> Suppressing the \fonttbl output makes autocad work almost perfectly (there are
> some other unrelated bugs in richedit code).
> Native riched20 didn't help either, as already told here.

What about riched30?

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #21 from [e-mail]  2009-04-25 04:48:41 ---
(In reply to comment #20)
> (In reply to comment #19)
> > @Austin : the log (with many printf() messages added) is already here, the
> > first attachment. As you can see, it shows nothing :-(
>
> Unless you changed every single warn/err/trace to a fixme, you're missing some
> output. Running with +richedit it much easier/faster.

I did. The added FIXMEs are *in addition* to +richedit logs... I did also
+richole and so on.
> > Suppressing the \fonttbl output makes autocad work almost perfectly (there are
> > some other unrelated bugs in richedit code).
> > Native riched20 didn't help either, as already told here.
>
> What about riched30?

Also tried, no way.
BTW, I found the cause of the bug.... I'll do some testing on native one and
then post the correct patch.... I guess some AutoCAD parts were written by some
brain-damaged people, indeed. They work just because of M$ richedit quirks.....

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

--- Comment #22 from [e-mail]  2009-04-25 12:57:07 ---
Fix sent to wine-patches.

Ciao

Max

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

[e-mail] changed:

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

--- Comment #23 from [e-mail]  2009-04-27 10:06:02 ---
Fix committed.

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

http://bugs.winehq.org/show_bug.cgi?id=14827

Alexandre Julliard  changed:

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

--- Comment #24 from Alexandre Julliard   2009-05-08 12:46:51 ---
Closing bugs fixed in 1.1.21.

[Bug 14827] Autocad 2005 : Multiline text edit crashes application

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=14827

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
      Fixed by SHA1|                            |6f173277281476b4872127bfaf2
                   |                            |f679d8f10ae96
                 CC|                            |[e-mail]
            Version|unspecified                 |1.1.2
          Component|-unknown                    |richedit