[Bug 25074] New: dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Summary: dotnet30 install fails with status 87 in wine > 1.3.4?
           Product: Wine
           Version: 1.3.5
          Platform: x86
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Doesn't happen in wine-1.3.4, but does in 1.3.5 and git.

First reported at http://code.google.com/p/winezeug/issues/detail?id=192

To see it in action, do 'winetricks dotnet30', then when the tray icon
is created, doubleclick it so the install window opens up again.
The failure happens in the first few seconds, the gui says there's
been an errror (but nothing suspicious happens in the console), and it
takes some time unwinding the install.  The installer then exits with nonzero
status, which in winetricks shows up at

Note: command 'wine $HOME/.cache/winetricks/dotnet30/dotnetfx3.exe' returned
status 87.  Aborting.

Bisecting is next.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #1 from Austin English  2010-11-08 03:15:13 CST ---
Probably a dupe of bug 25003.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Dan Kegel  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download, Installer,
                   |                            |regression
          Component|-unknown                    |crypt32

--- Comment #2 from Dan Kegel  2010-11-09 13:22:51 CST ---
The culprit seems to be

http://source.winehq.org/git/wine.git/?a=commit;h=966d722752b659a12ffa355a1e...
crypt32: Improve error checking for the base policy.
author    Juan Lang 
     Tue, 5 Oct 2010 01:16:16 +0000 (18:16 -0700)

That patch doesn't revert cleanly with current git, but I did verify
that
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  git reset --hard HEAD^
  ./configure 
  make
generated a wine that could install dotnet30, but
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  ./configure 
  make
generated one that couldn't.

( For completeness, here's how I bisected: I created bisect.sh containing
 #!/bin/sh
 CFLAGS="-g -O0" ./configure --disable-tests || exit 125
 make -j16 || exit 125
 rm -rf $HOME/.wine || exit 125
 WINE=$HOME/wine-git/wine sh ~/winezeug/winetricks -q gecko dotnet30
and ran the commands
  git bisect start
  git bisect wine-1.3.4 good
  git bisect good wine-1.3.4 
  git bisect bad wine-1.3.5
  git bisect run sh bisect.sh 
)

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #3 from Juan Lang  2010-11-09 18:20:10 CST ---
The bug is more likely to be in wintrust than in crypt32, but I'll defer
changing the component until we get more evidence.  Please attach a
+wintrust,+crypt log.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Juan Lang  changed:

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

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #4 from Louis Lenders  2010-11-10 11:23:24 CST ---
Created an attachment (id=31831)
 --> (http://bugs.winehq.org/attachment.cgi?id=31831)
+crypt,+wintrust

This is the log I get. It's pretty short .....

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #5 from Juan Lang  2010-11-10 11:30:54 CST ---
I got something similar.  The function modified by that commit isn't called in
this log.  Dan, are you sure the bisect is correct?

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #6 from Dan Kegel  2010-11-10 12:42:48 CST ---
Pretty sure.  I did double-check as noted above:

I did verify that
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  git reset --hard HEAD^
  ./configure 
  make
generated a wine that could install dotnet30, but
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  ./configure 
  make
generated one that couldn't.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #7 from Juan Lang  2010-11-10 14:20:10 CST ---
Created an attachment (id=31833)
 --> (http://bugs.winehq.org/attachment.cgi?id=31833)
Patch: Revert 966d722752b659a12ffa355a1e559f94907cd66d

Does this patch help, then?  It'll cause test failures, but it'll at least
confirm whether the identified commit is to blame.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #8 from Dan Kegel  2010-11-10 23:37:29 CST ---
Nope.  Maybe something else broke in the meantime.

I verified again that
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  git reset --hard HEAD^
  ./configure 
  make
  rm -rf $HOME/.wine
  WINE=$HOME/wine-git/wine sh ~/winezeug/winetricks -q gecko dotnet30

worked, and

git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  ./configure 
  make
  rm -rf $HOME/.wine 
  WINE=$HOME/wine-git/wine sh ~/winezeug/winetricks -q gecko dotnet30

generated the familiar error 87.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Juan Lang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|crypt32                     |-unknown

--- Comment #9 from Juan Lang  2010-11-11 12:09:25 CST ---
We'll have to find something else to blame for the time being, then.  At least,
I can't easily debug the current breakage if the function isn't called anymore.
 Probably there were multiple regressions.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #10 from Juan Lang  2010-11-12 07:33:56 CST ---
Could someone attach a +crypt,+wintrust log from
966d722752b659a12ffa355a1e559f94907cd66d ?

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #11 from Austin English  2010-11-12 13:30:36 CST ---
Created an attachment (id=31882)
 --> (http://bugs.winehq.org/attachment.cgi?id=31882)
+crypt,+wintrust at 966d722752b659a12ffa355a1e559f94907cd66d

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Juan Lang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|-unknown                    |wintrust

--- Comment #12 from Juan Lang  2010-11-12 16:52:55 CST ---
Thanks.  From the log:
trace:wintrust:WINTRUST_DefaultVerifyAndClose returning 800b0101
trace:wintrust:WinVerifyTrust returning 800b0101

Basically, the time to verify needs to come from the cat file.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Juan Lang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|wintrust                    |-unknown

--- Comment #13 from Juan Lang  2010-11-13 07:51:57 CST ---
Hm.  Except that SoftpubLoadSignature is called:
trace:wintrust:SoftpubLoadSignature (0x13fc38)
and it's already been fixed to get the time to sign from the message, rather
than from the file.  In other words, I believe the wintrust regression may
already have been fixed.

You may need to carry forward commit 68fceb5e5b4c345dd6d7f77d2d250f1e8a779007
and perform a regression test with that one applied to see where else it broke.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #14 from Juan Lang  2010-11-15 17:45:42 CST ---
I just confirmed the regression test, and my suspicion that the commit
identified has already been fixed.

I.e.,
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  ./configure 
  make
produces a wine that fails to install dotnet30 (though with error 67 for me,
not 87 as reported.)  And,
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  git reset --hard HEAD~1
  ./configure 
  make
produces a wine that installs dotnet30.  However,
  git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
  git cherry-pick 68fceb5e5b4c345dd6d7f77d2d250f1e8a779007
  ./configure 
  make
also produces a wine that installs dotnet30.

So, the remaining regression is elsewhere, and a new regression test is needed.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Austin English  changed:

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

--- Comment #15 from Austin English  2010-11-15 23:00:44 CST ---
(In reply to comment #14)
> I just confirmed the regression test, and my suspicion that the commit
> identified has already been fixed.
>
> I.e.,
>   git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
>   ./configure 
>   make
> produces a wine that fails to install dotnet30 (though with error 67 for me,
> not 87 as reported.)  And,
>   git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
>   git reset --hard HEAD~1
>   ./configure 
>   make
> produces a wine that installs dotnet30.  However,
>   git reset --hard 966d722752b659a12ffa355a1e559f94907cd66d
>   git cherry-pick 68fceb5e5b4c345dd6d7f77d2d250f1e8a779007
>   ./configure 
>   make
> also produces a wine that installs dotnet30.
>
> So, the remaining regression is elsewhere, and a new regression test is needed.

austin@aw21 ~/wine-git $ git bisect bad
ca7cd1bf89d057d9c48896b70c441fad941e6ee0 is the first bad commit
commit ca7cd1bf89d057d9c48896b70c441fad941e6ee0
Author: Hans Leidekker 
Date:   Wed Oct 6 15:44:08 2010 +0200

msi: Validate packages based on supported version, platform and languages.

:040000 040000 1ef576e28c83ea59db6043d13d6921f475e26b9d
9d1df66c16e2c640d0a24677ec122f1ac773421c M      dlls

Reverting that commit on HEAD fails, but I reverted it over itself (after
cherry-picking 68fceb5e5b4c345dd6d7f77d2d250f1e8a779007) and that succeeds.

Unless there's a third regression somewhere, fixing this should fix dotnet30.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Austin English  changed:

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

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #16 from Hans Leidekker  2010-11-16 03:00:50 CST ---
Created an attachment (id=31973)
 --> (http://bugs.winehq.org/attachment.cgi?id=31973)
msi: Fix handling of an empty language id list in msi_parse_summary.

This patch sould fix that particular regression.
> Unless there's a third regression somewhere, fixing this should fix dotnet30.

I'm afraid there is, and even more. It needs a builtin override for
aspnet_regiis.exe and then fails again in a custom action that runs
this command:

C:\windows\Microsoft.NET\Framework\v3.0\Windows Communication
Foundation\ServiceModelReg.exe /u /x /y /v

which seems to be caused by missing WMI interfaces.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Louis Lenders  changed:

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

--- Comment #17 from Louis Lenders  2010-11-16 11:34:13 CST ---
(In reply to comment #16)
> Created an attachment (id=31973)

--> (http://bugs.winehq.org/attachment.cgi?id=31973) [details]
> msi: Fix handling of an empty language id list in msi_parse_summary.
>
> This patch sould fix that particular regression.
> > Unless there's a third regression somewhere, fixing this should fix dotnet30.
>
> I'm afraid there is, and even more.

Hi Hans, I applied the patch, and after 'winetricks dotnet30' the installer
finishes just fine for me. I don;t see any other regressions atm. Did the
installer crash for you?

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

A Wine user  changed:

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

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #18 from Hans Leidekker  2010-11-16 11:51:59 CST ---
> Hi Hans, I applied the patch, and after 'winetricks dotnet30' the installer
> finishes just fine for me. I don;t see any other regressions atm. Did the
> installer crash for you?

It didn't crash but it did return an error.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #19 from Louis Lenders  2010-11-16 11:57:51 CST ---
(In reply to comment #18)
> > Hi Hans, I applied the patch, and after 'winetricks dotnet30' the installer
> > finishes just fine for me. I don;t see any other regressions atm. Did the
> > installer crash for you?
>
> It didn't crash but it did return an error.

For me it didn;t return an error, and finished just fine. Anyone else could try
this?

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #20 from A Wine user  2010-11-16 12:16:23 CST ---
With the patch applied, winetricks dotnet30 gave me "an error occurred during
setup. Installed files have been removed from your computer" and a prompt to
send an error report to MS.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

--- Comment #21 from Hans Leidekker  2010-11-18 02:12:48 CST ---
The regression described in comment 15 is fixed by commit
a74e025335b107926bf9062ef6ac0fa120613261.

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Austin English  changed:

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

--- Comment #22 from Austin English  2010-11-19 01:38:20 CST ---
(In reply to comment #21)
> The regression described in comment 15 is fixed by commit
> a74e025335b107926bf9062ef6ac0fa120613261.

Works great, thanks!

[Bug 25074] dotnet30 install fails with status 87 in wine > 1.3.4?

http://bugs.winehq.org/show_bug.cgi?id=25074

Alexandre Julliard  changed:

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

--- Comment #23 from Alexandre Julliard  2010-11-26 13:13:51 CST ---
Closing bugs fixed in 1.3.8.