[Bug 11488] New: corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Summary: corefont installers hanging
           Product: Wine
           Version: CVS/GIT
          Platform: Other
               URL: http://downloads.sourceforge.net/corefonts/arial32.exe
        OS/Version: other
            Status: NEW
          Keywords: download, Installer, regression
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

With today's git, the ms corefont installers seem to be hanging on exit.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #1 from Dan Kegel   2008-02-06 09:35:48 ---
With this morning's git, it's worse: they hang earlier
with the following deadlock:

err:ntdll:RtlpWaitForCriticalSection section 0x12cb74 "animate.c:
ANIMATE_INFO*->cs" wait timed out in thread 0018, blocked by 0019, retrying (60
sec)

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #2 from Dan Kegel   2008-02-07 08:05:15 ---
Rob's two patches 
http://www.winehq.org/pipermail/wine-patches/2008-February/049721.html
http://www.winehq.org/pipermail/wine-patches/2008-February/049764.html
seem to fix the animation deadlock, but the corefont installer
still has the hang I opened this bug for.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #3 from Austin English   2008-02-07 11:19:10 ---
This also affects Speed Commander 12. Regression testing points to:

austin@austin-desktop:~/wine-git$ git bisect good
0edae1c9eaf7d49fb524e3a392b7ed2949c7edce is first bad commit
commit 0edae1c9eaf7d49fb524e3a392b7ed2949c7edce
Author: Reece H. Dunn 
Date:   Sat Jan 26 10:34:45 2008 +0000
    cabinet: Make the fdi tests pass on Vista.

Dan, can you see if reverting that patch fixes your problems as well?

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Austin English  changed:

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

--- Comment #4 from Austin English   2008-02-07 11:19:26 ---
Forgot to cc author...

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #5 from Reece Dunn   2008-02-07 11:49:00 ---
(In reply to comment #3)
> This also affects Speed Commander 12. Regression testing points to:
>
> austin@austin-desktop:~/wine-git$ git bisect good
> 0edae1c9eaf7d49fb524e3a392b7ed2949c7edce is first bad commit
> commit 0edae1c9eaf7d49fb524e3a392b7ed2949c7edce
> Author: Reece H. Dunn 
> Date:   Sat Jan 26 10:34:45 2008 +0000
>     cabinet: Make the fdi tests pass on Vista.
>
> Dan, can you see if reverting that patch fixes your problems as well?

That patch only affects the tests (dlls/cabinet/tests/fdi.c), so I doubt that
it is this patch that is causing the regression.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #6 from Lei Zhang   2008-02-07 13:00:55 ---
Weird, I'd help with the regression test, but I'm not seeing this problem at
all.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #7 from Reece Dunn   2008-02-07 13:15:47 ---
I've just used wget to download the executable and run it with a build of the
latest Wine HEAD. I can't reproduce the hang either. I'm using Ubuntu Gutsy
Gibbon.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #8 from Austin English   2008-02-07 14:52:19 ---
(In reply to comment #5)
> (In reply to comment #3)
> > This also affects Speed Commander 12. Regression testing points to:
> >
> > austin@austin-desktop:~/wine-git$ git bisect good
> > 0edae1c9eaf7d49fb524e3a392b7ed2949c7edce is first bad commit
> > commit 0edae1c9eaf7d49fb524e3a392b7ed2949c7edce
> > Author: Reece H. Dunn 
> > Date:   Sat Jan 26 10:34:45 2008 +0000
> >     cabinet: Make the fdi tests pass on Vista.
> >
> > Dan, can you see if reverting that patch fixes your problems as well?
>
> That patch only affects the tests (dlls/cabinet/tests/fdi.c), so I doubt that
> it is this patch that is causing the regression.

Damn, you're right. I didn't think that sounded right, ran git revert, and it
fixed it, but seems I tainted the results somewhat still. I'm not at that
machine right now, I'll rerun the test in an hour or so.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #9 from Dan Kegel   2008-02-07 23:24:55 ---
I finally grabbed git from today (Feb 7th) and rebuilt,
and the problem is still happening.  Once one font installer
did terminate ok, but five other times it hung on exit.

Here's the last 25 or so lines of the log up to the hang with +relay.
Highlights:
0016:Ret  user32.EndDialog() retval=00000001 ret=0100307f
...
0016:Call KERNEL32.WaitForSingleObject(00000068,ffffffff) ret=7e9eb373
001a:Ret  KERNEL32.WaitForSingleObject() retval=00000102 ret=7e9ed275
...
001a:Call user32.SendMessageW(00040024,00000138,000001d8,0003002e) ret=7e9ed2b4
[hang here]

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #10 from Dan Kegel   2008-02-07 23:25:33 ---
Created an attachment (id=10672)
 --> (http://bugs.winehq.org/attachment.cgi?id=10672)
+relay log leading up to the hang

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #11 from Dan Kegel   2008-02-07 23:41:55 ---
If you're doing a git bisect, you probably have
to repeat running the app several times to
give the bug a chance to show itself...

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #12 from Austin English   2008-02-08 17:00:12 ---
(In reply to comment #11)
> If you're doing a git bisect, you probably have
> to repeat running the app several times to
> give the bug a chance to show itself...

It's pretty consistent for me, though we could have two separate bugs. Ran
regression test again, and if the install succeeded 3 times, I considered it
good, if it hung, failed. Got this patch:

6d59ddf67b6beb8862a4b784c9a52e8a34b5c13d - 
kernel32: Don't fail unconditionally in MoveFile for directories with flag
MOVEFILE_REPLACE_EXISTING.
http://source.winehq.org/git/wine.git/?a=commitdiff;h=6d59ddf67b6beb8862a4b7...

Reverting it in wine 0.9.55 -> happy installer. Care to give that a try Dan?

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #13 from Dan Kegel   2008-02-09 11:16:55 ---
On one machine, I get no hangs anyway.

On another machine, reverting that patch or not, I still get the
hang I described.  I'll try building from a clean tree just in case.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #14 from Austin English   2008-02-09 13:41:32 ---
(In reply to comment #13)
> On one machine, I get no hangs anyway.
>
> On another machine, reverting that patch or not, I still get the
> hang I described.  I'll try building from a clean tree just in case.

Works fine for me in gutsy, only hangs in my feisty VM. Could you try supreme
commander's installer. We may have two different bugs, but the sound
similar...For me, SC installs fine in 0.9.54 (well, the installer has other
bugs, but it completes). In current git/after that kernel32 patch, it gets to
the point of copying files and hangs, not copying anything. Never succeeds
after that. I'd appreciate you testing it so we can see if it's a separate bug
or not.

http://www.speedproject.de/enu/speedcommander/

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #15 from Dan Kegel   2008-02-09 16:13:12 ---
Fresh from git, on my bad system (Ubuntu Gutsy), still hangs.

I tried speedcommander 12.1's installer; it doesn't hang
( though, bizarrely, it installs "Default" instead of the app ).

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #16 from Dan Kegel   2008-02-10 02:52:30 ---
For me,
  git checkout 43083236c428e968a7fa1c54470a67cd89674034
  ./configure --prefix=/usr/local/wine-20080207
  make depend && make
  ./wine ../winetrickscache/arial32.exe
reproduces the error, i.e. arial32 hangs after
the progress window showing the copying files goes away.

I chose that because it was said to be the parent of
6d59ddf67b6beb8862a4b784c9a52e8a34b5c13d.  And indeed, then
doing a git checkout of that revision and then building
shows only kernel32/path.c was rebuilt.  So I think
your bisect might have been wrong, the real problem
is earlier.  Unless we're chasing different problems.

BTW, I'm having trouble doing a bisect:
$ git checkout wine-0.9.54
HEAD is now at 8f954cc... Release 0.9.54.
$ git bisect start
fatal: ref HEAD is not a symbolic ref
Bad HEAD - I need a symbolic ref

gaah!  Any tips?  This is a fairly fresh source tree,
the only funky thing I've done with it is do a few 'git checkout nnnnn's.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #17 from Reece Dunn   2008-02-10 04:56:10 ---
(In reply to comment #16)
> For me,
>   git checkout 43083236c428e968a7fa1c54470a67cd89674034
>   ./configure --prefix=/usr/local/wine-20080207
>   make depend && make
>   ./wine ../winetrickscache/arial32.exe
> reproduces the error, i.e. arial32 hangs after
> the progress window showing the copying files goes away.

Dan, your error looks like the wizard is not terminating its message loop. I
would suspect it is not being sent a WM_QUIT message (via PostQWuitMessage) for
some reason.

0016:Ret  user32.EndDialog() retval=00000001 ret=0100307f
...
001a:Call user32.SendMessageW(00040024,00000138,000001d8,0003002e) ret=7e9ed2b4
[hang here]

The thing that interests me about this is that the message id (138, or 8a in
hex - note the trace there would be more useful to have the message in hex as
that is what they are in winuser.h) is not in winuser.h. Also, the
dlls/comctl32/tests/propsheet.c tests are very minimal.
> I chose that because it was said to be the parent of
> 6d59ddf67b6beb8862a4b784c9a52e8a34b5c13d.  And indeed, then
> doing a git checkout of that revision and then building
> shows only kernel32/path.c was rebuilt.  So I think
> your bisect might have been wrong, the real problem
> is earlier.  Unless we're chasing different problems.

I suspect that this is the case. That change is doing an NtOpenFile/NtCloseFile
for a directory. From what Austin said:
> In current git/after that kernel32 patch, it gets to the
> point of copying files and hangs, not copying anything.
> Never succeeds after that.

The installer is hanging before it is copying files, not after it is completed.
For Austin's issue, that patch looks valid. Does anyone know what the test
coverage is like for this, w.r.t. the MOVEFILE_REPLACE_EXISTING flag for
MoveFile on directories that don't exist, exist and exist, but with different
permissions (e.g. something owned by root)?
> BTW, I'm having trouble doing a bisect:
> $ git checkout wine-0.9.54
> HEAD is now at 8f954cc... Release 0.9.54.
> $ git bisect start
> fatal: ref HEAD is not a symbolic ref
> Bad HEAD - I need a symbolic ref
>
> gaah!  Any tips?  This is a fairly fresh source tree,
> the only funky thing I've done with it is do a few 'git checkout nnnnn's.

You need to do something like:
$ git checkout master
$ git bisect start
$ git bisect bad
$ git checkout wine-0.9.54
$ git bisect good

Bisect start doesn't like being on tags or at a random point.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #18 from Dmitry Timoshkov   2008-02-10 07:13:38 ---
(In reply to comment #17)
> 001a:Call user32.SendMessageW(00040024,00000138,000001d8,0003002e) ret=7e9ed2b4
> [hang here]
> The thing that interests me about this is that the message id (138, or 8a in
> hex - note the trace there would be more useful to have the message in hex as
> that is what they are in winuser.h) is not in winuser.h.

All the numbers in a relay trace are in hex. 138 above is in hex.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #19 from Reece Dunn   2008-02-10 08:11:55 ---
(In reply to comment #18)
> (In reply to comment #17)
> > 001a:Call user32.SendMessageW(00040024,00000138,000001d8,0003002e) ret=7e9ed2b4
> > [hang here]
> > The thing that interests me about this is that the message id (138, or 8a in
> > hex - note the trace there would be more useful to have the message in hex as
> > that is what they are in winuser.h) is not in winuser.h.
>
> All the numbers in a relay trace are in hex. 138 above is in hex.

Ok. Thanks. Message 138 is WM_CTLCOLORSTATIC.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #20 from Dmitry Timoshkov   2008-02-10 08:19:59 ---
The regression might be related to recent animation control fixes from Rob.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #21 from Dan Kegel   2008-02-11 00:34:11 ---
Yup.  I figured out how to do git bisect, but
only got garbage results out of them, probably
because I wasn't careful enough.

git reset --hard  7bc9e425f166a9f7473f1526ff582c1e18ccf6e3; make
is good, and 
git reset --hard  53c6cdf918e10305f19e74e003fc4d2311ed556f; make
is bad, i.e. hangs after the copy progress window goes away.
So the bad commit is

commit 53c6cdf918e10305f19e74e003fc4d2311ed556f
Author: Rob Shearman 
Date:   Thu Jan 31 14:47:03 2008 +0000
comctl32: Get the background colour in ANIMATE_PaintFrame, instead of just
in WM_PAINT.

(Either way, git-rev-list --max-count=1 origin
shows 1991e9f9406ae8f72663aace7e2cfa6afc76e748, i.e. wine-0.9.55,
so my tree isn't disconnected, if that's what you're worried about, Rob.)

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Dan Kegel  changed:

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

--- Comment #22 from Dan Kegel   2008-02-11 00:35:03 ---
*** Bug 11534 has been marked as a duplicate of this bug. ***

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Dan Kegel  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |major
          Component|-unknown                    |comctl32

--- Comment #23 from Dan Kegel   2008-02-11 00:37:30 ---
Bug 11534 says this affects the .NET Framework 2.0 installer, too.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #24 from Austin English   2008-02-11 13:58:42 ---
Looks like we've got different bugs, I'll file a separate one for it.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Dan Kegel  changed:

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

--- Comment #25 from Dan Kegel   2008-02-12 09:03:45 ---
Works with yesterday's git, probably fixed by
http://www.winehq.org/pipermail/wine-cvs/2008-February/040176.html
Thanks, Alexandre!

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

James Hawkins  changed:

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

--- Comment #26 from James Hawkins   2008-02-17 12:11:01 ---
*** Bug 11623 has been marked as a duplicate of this bug. ***

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

--- Comment #27 from Vitaliy Margolen   2008-02-17 18:06:24 ---
*** Bug 11623 has been marked as a duplicate of this bug. ***

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Dan Kegel  changed:

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

--- Comment #28 from Dan Kegel   2008-02-22 11:40:16 ---
0.9.56 released, so closing all bugs marked as RESOLVED FIXED.

[Bug 11488] corefont installers hanging

http://bugs.winehq.org/show_bug.cgi?id=11488

Austin English  changed:

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