http://bugs.winehq.org/show_bug.cgi?id=29949
Bug #: 29949 Summary: PAF (Personal Ancestral File) help viewer unhandled page fault Product: Wine Version: 1.4-rc4 Platform: x86 OS/Version: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: rmriches@ieee.org Classification: Unclassified
Created attachment 38982 --> http://bugs.winehq.org/attachment.cgi?id=38982 stdout and stderr from a run that shows the unhandled page fault
This is a regression. This problem does not exist in 1.4-rc3. I plan to run a regression test to find the exact commit. The application is downloadable.
Start from a fresh WINE prefix area with no pre-existing WINE processes running. When starting the application, it presents a 'Welcome ...' dialog. Click on 'New'. Enter any filename, click 'Save'. Click 'Cancel' on the next two dialogs. This puts you in the family view with no pre-existing data. I get the same results even if I use pre-existing data, but it's easier to bring up without pre-existing data.
Click on the yellow question mark in the icon bar near the top of the window. This brings up the help viewer.
In the left-hand pane of the help viewer, click on chapter titles starting at the top. Click at least one per second. At some point between about the seventh entry and the bottom of the list, you will see a dialog saying the application needs to close. The console will show an unhandled page fault. I'll attach console output.
The visible symptoms here look very similar to some old bugs seen in 1.3.12-1.3.15, 0.9.57-0.9.60, 0.9.50, 0.9.52, 0.9.39, and 0.9.37. Bug number 25873 is a good example.
http://bugs.winehq.org/show_bug.cgi?id=29949
Dmitry Timoshkov dmitry@baikal.ru changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |regression
--- Comment #1 from Dmitry Timoshkov dmitry@baikal.ru 2012-02-19 22:29:35 CST --- Please post the results of your regression testing: http://wiki.winehq.org/RegressionTesting
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #2 from Robert Riches rmriches@ieee.org 2012-02-19 22:47:02 CST --- Here's what I got from a regression bisect test. It doesn't seem to make sense to me that updating a German translation could have a significant effect on a program installed and used in English. However, maybe there could be a second-order effect hiding somewhere. This isn't my first regression bisect run, and I think I did it correctly, but then again, I'm not infallible. Anyway, this is what I got:
bc058788de460e774ef8dde4f9b64f5f86e9e36a is the first bad commit commit bc058788de460e774ef8dde4f9b64f5f86e9e36a Author: André Hentschel nerv@dawncrow.de Date: Sun Feb 12 19:37:01 2012 +0100
po: Update German translation.
:040000 040000 0e7f68cceeb351ae398bde2c2f3dc957cae0967d 24bcdd42c2fec8d5e3258c86e4eeac1d89a4d40e M po
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #3 from Dmitry Timoshkov dmitry@baikal.ru 2012-02-19 23:02:39 CST --- Does reverting that patch help?
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #4 from Robert Riches rmriches@ieee.org 2012-02-19 23:16:26 CST --- It appears the act of reverting the patch failed. I followed the instructions to reset the bisect and revert the patch. This is the command I used to revert the patch:
git show bc058788de460e774ef8dde4f9b64f5f86e9e36a | patch -p1 -R
Here's what I got:
patching file po/de.po Hunk #1 FAILED at 5. Hunk #3 succeeded at 9667 with fuzz 1. 1 out of 5 hunks FAILED -- saving rejects to file po/de.po.rej
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #5 from Dmitry Timoshkov dmitry@baikal.ru 2012-02-20 00:12:33 CST --- Try instructions at http://wiki.winehq.org/RegressionTesting#head-10b19f220a1073e436fa2cf55e3dad...
http://bugs.winehq.org/show_bug.cgi?id=29949
Michael Stefaniuc mstefani@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |mstefani@redhat.com
--- Comment #6 from Michael Stefaniuc mstefani@redhat.com 2012-02-20 07:01:48 CST --- It is highly unlikely that the German translation patch is the issue. It looks more like a regression testing error. To cross check in case of issues to revert a patch you can run: # without commit git checkout bc058788de460e774ef8dde4f9b64f5f86e9e36a^ # with commit git checkout bc058788de460e774ef8dde4f9b64f5f86e9e36a
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #7 from Robert Riches rmriches@ieee.org 2012-02-20 09:44:18 CST --- Relative to comment #5, those are the instructions I had been following.
Relative to comment #6, the pre-commit version shows the bad symptoms, so something must have gone wrong with the regression run. I'll try again when I get an opportunity--later today if possible.
I would think it would be good for the cross-check instructions in comment #6 to be included in the instructions page on running a regression, right after 'Resetting the bisect'. Is there a procedure to request such a change? I have Bugzilla and AppDB access but probably not Wiki write access.
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #8 from Robert Riches rmriches@ieee.org 2012-02-20 12:41:45 CST --- Ran the regression again from the start. This is what I got from the regression run this time:
8e1dbebdb0e7b834fad53b4ea3d3cd349c34fbec is the first bad commit commit 8e1dbebdb0e7b834fad53b4ea3d3cd349c34fbec Author: Alexandre Julliard julliard@winehq.org Date: Mon Feb 13 12:28:32 2012 +0100
hhctrl.ocx: Fix a buffer allocation size.
:040000 040000 9c3402bf12e5f8ae6b2524be350a5307b8adabf3 6f075ef166278ca3c7e9e5daa7b93cf536771fd6 M dlls
However, reverting that patch did not fix the symptoms. The symptoms do change a little from run to run, even with the same compiled code; namely, sometimes it's an unhandled page fault that crashes WINE, and sometimes WINE just hangs.
The somewhat non-deterministic symptoms and such are causing me to suspect a memory allocation or reference issue of the type Valgrind is made to solve. Is WINE normally Valgrind-clean? Would it make sense to try using Valgrind to find this thing?
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #9 from Robert Riches rmriches@ieee.org 2012-02-26 12:11:33 CST --- Problem behavior still seen in 1.4-rc5.
Any suggestions to narrow this down that would be more likely to yield useful results than the two regression test runs already done?
http://bugs.winehq.org/show_bug.cgi?id=29949
Frédéric Delanoy frederic.delanoy@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |frederic.delanoy@gmail.com Regression SHA1| |8e1dbebdb0e7b834fad53b4ea3d | |3cd349c34fbec
http://bugs.winehq.org/show_bug.cgi?id=29949
Frédéric Delanoy frederic.delanoy@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Regression SHA1|8e1dbebdb0e7b834fad53b4ea3d | |3cd349c34fbec |
http://bugs.winehq.org/show_bug.cgi?id=29949
Jacek Caban jacek@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jacek@codeweavers.com
--- Comment #10 from Jacek Caban jacek@codeweavers.com 2012-02-27 04:41:36 CST --- Does the behavior changes when ran with warn+heap? If it still crashes in Gecko, it may be worth trying to get proper backtrace with Gecko debug build.
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #11 from Robert Riches rmriches@ieee.org 2012-02-28 23:21:10 CST --- Created attachment 39133 --> http://bugs.winehq.org/attachment.cgi?id=39133 log file with warn+heap and crash while exiting application
This is a plain text log file from a run with 1.4-rc5 with WINEDEBUG set to warn+heap. (Tried to gzip it, but that didn't work.) In two attempts, the unhandled page fault and crash did not happen until exiting the application after going through the sequence of clicking on the help viewer chapter titles and then closing the help viewer (window manager close button in title bar).
Jacek, was that what you were looking for?
Or, do I need to use a special build of Gecko?
http://bugs.winehq.org/show_bug.cgi?id=29949
Jacek Caban jacek@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |NEW Ever Confirmed|0 |1
--- Comment #12 from Jacek Caban jacek@codeweavers.com 2012-02-29 04:15:19 CST --- This patch should help:
http://source.winehq.org/patches/data/83971
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #13 from Michael Stefaniuc mstefani@redhat.com 2012-02-29 14:15:35 CST --- Patch committed http://source.winehq.org/git/wine.git/?a=commit;h=981d18464f80ca87c4d1aaf394...
http://bugs.winehq.org/show_bug.cgi?id=29949
Jerome Leclanche adys.wh@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Fixed by SHA1| |981d18464f80ca87c4d1aaf3942 | |26c6c3ffda8d8 Status|NEW |RESOLVED CC| |adys.wh@gmail.com Resolution| |FIXED
--- Comment #14 from Jerome Leclanche adys.wh@gmail.com 2012-02-29 14:22:51 CST --- Fixed; please reopen if issue persists.
http://bugs.winehq.org/show_bug.cgi?id=29949
--- Comment #15 from Robert Riches rmriches@ieee.org 2012-02-29 20:00:28 CST --- Appears to work well on my side. Thank you!
http://bugs.winehq.org/show_bug.cgi?id=29949
Alexandre Julliard julliard@winehq.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #16 from Alexandre Julliard julliard@winehq.org 2012-03-02 13:02:26 CST --- Closing bugs fixed in 1.4-rc6.
-
wine-bugs＠winehq.org