[Bug 15377] New: 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Summary: 1.1.5 breaks d3d game that worked in 1.1.4
           Product: Wine
           Version: 1.1.5
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: major
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]

broke both command and conquer & command and conquer: red alert in wine 1.1.5;
both worked in 1.1.4

Regression Testing

d381c7cbf4d463189f479a0b3986f6767aeb8dba is first bad commit
commit d381c7cbf4d463189f479a0b3986f6767aeb8dba
Author: Alexandre Julliard 
Date:   Fri Sep 19 18:05:36 2008 +0200

Release 1.1.5.

:100644 100644 7a2c61113334222ab571b9cce093dd080fa825bf
982ac735f37215caf4e819c8afb6a8a2f0341585 M      ANNOUNCE
:100644 100644 cd289fc7fdb8187487084f08312e3042c1c7cb44
41c2c19075c1b563e1152512ed077b9895bd1f26 M      VERSION
:100755 100755 e898943c5e9326274904e3ecc42884b99456e288
ff2c11d3ae4e0afd80c9b62281c510521cef9e87 M      configure

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #1 from James Hawkins   2008-09-22 20:06:54 ---
You didn't do the regression test correctly.  There is no way that commit broke
your game.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression

--- Comment #2 from Austin English   2008-09-22 20:07:25 ---
Rerun the regression test. That couldn't have been the cause.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #3 from Ben Trusty   2008-09-22 20:38:06 ---
(In reply to comment #2)
> Rerun the regression test. That couldn't have been the cause.

i did this,
http://wiki.winehq.org/RegressionTesting

git bisect start
git bisect good 1.1.4
git bisect bad 1.1.5

<start>
CC="ccache gcc" ./configure --verbose && make clean && make

{skipping listing part about backing up .wine directory and reinstalling game}
./wine "C:\Program Files\Program Name\program.exe"

{bug did not occur}
git bisect good

<goto start>, until it said that there were no more, e.g.  
0 revisions left to test

and it gave me the information that i posted in the bug

...

what did i do wrong?
it somewhat hard to just 're-do' it quickly. as it took about 4-6 hours to
recompile the times it took to get it done.

suggestions?

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #4 from James Hawkins   2008-09-22 20:43:40 ---
(In reply to comment #3)
> <goto start>, until it said that there were no more, e.g.  
> 0 revisions left to test
>
> what did i do wrong?

Quoting from the regression testing site:

"If you see:

Bisecting: 0 revisions left to test after this

THEN YOU ARE NOT DONE YET!"

If you haven't done a git bisect reset, then you should still be able to finish
the test with one more iteration.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #5 from Ben Trusty   2008-09-22 21:16:13 ---
(In reply to comment #4)
> (In reply to comment #3)
> > <goto start>, until it said that there were no more, e.g.  
> > 0 revisions left to test
> >
> > what did i do wrong?
>
> Quoting from the regression testing site:
>
> "If you see:
>
> Bisecting: 0 revisions left to test after this
>
> THEN YOU ARE NOT DONE YET!"
>
> If you haven't done a git bisect reset, then you should still be able to finish
> the test with one more iteration.

git bisect reset
<replies>
Switched to branch "master"

CC="ccache gcc" ./configure --verbose && make clean && make

run program again
doesnt work anymore

idea on what to do next?

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #6 from James Hawkins   2008-09-22 21:19:35 ---
You've gotta slow down and read more carefully.  I said if you *haven't* done a
git bisect reset, then you're almost done, but now you've reset the bisect and
you have to do it all over again.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #7 from Ben Trusty   2008-09-22 21:25:58 ---
(In reply to comment #6)
> You've gotta slow down and read more carefully.  I said if you *haven't* done a
> git bisect reset, then you're almost done, but now you've reset the bisect and
> you have to do it all over again.

so wait, your telling me that i have to...

1 git bisect start
2 git bisect good wine-1.1.4
3 git bisect bad wine-1.1.5
4 CC="ccache gcc" ./configure --verbose && make clean && make
5 ./wine "C:\Program Files\Program Name\program.exe"

if program doesnt work
 git bisect good
if program doesnt work
 git bitset bad

continue 4 & 5 until 0 remaining

git bisect reset

continue 4 & 5 until 0 remaining again?

then what, when i get to 0 again? and do git bisect reset again?
seems like im getting into an infinite loop.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #8 from James Hawkins   2008-09-22 21:32:24 ---
No, you don't use git bisect reset *at all* until the entire process is
completely finished.  I only brought it up because you could've saved yourself
a lot of time.  Just follow the directions on that site...it's not that hard.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #9 from Ben Trusty   2008-09-22 21:42:28 ---
(In reply to comment #8)
> No, you don't use git bisect reset *at all* until the entire process is
> completely finished.  I only brought it up because you could've saved yourself
> a lot of time.  Just follow the directions on that site...it's not that hard.

i just listed what i thought the site was the order of operations

the site http://wiki.winehq.org/RegressionTesting
has a whole list of "what ifs" added into the order of operations
if there is an easier list, that would be helpful, as what i just posted was
the order of what i was supposed to do.

b/c i did initially what i thought was done and got to a point that said
>

THEN YOU ARE NOT DONE YET!

Once that patch is identified, post this info into bugzilla in the bug report
(assuming one is opened, if not, open one). Also be sure to add the author of
the patch to the CC. This information will allow the developers to fix the
problem much quicker.
>

post this info into bugzilla in the bug report (assuming one is opened, if not,
open one)

so i did that, which oddly there is something AFTER that (e.g. git bisect
reset), something i did AFTER i had completed and had some feedback from the
process (e.g. what i posted that the problem was which was causing the game to
not function in 1.1.5

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #10 from Austin English   2008-09-22 22:49:35 ---
Assuming you've got git compiling fine:

1. $ git bisect reset
2. $ git bisect start
3. $ git bisect good wine-1.1.4
4. $ git bisect bad wine-1.1.5
5. $ CC="ccache gcc" ./configure && make clean && make depend && make
6. $ ~/wine-git/wine your_program.exe
7. If good A) $ git bisect good
   If bad B) $ git bisect bad

Repeat until you see something similar to:

a460a2df43aa8eae10b1db028c6020829b009c54 is first bad commit
commit a460a2df43aa8eae10b1db028c6020829b009c54
Author: Stefan Doesinger 
Date:   Sat Jun 9 14:27:41 2007 +0200
wined3d: Store the gl information in a per adapter structure and initialize it
only once.
:040000 040000 d8ae35832fcdbca8de07acae73b9e21564ced413
1720cc38fb598110071c9ee4f21f8f61b6f764c3 M      dlls

If it shows a release, like you got, then there's something wrong...likely a
race condition, etc.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #11 from James Hawkins   2008-09-22 23:51:49 ---
(In reply to comment #10)
> Assuming you've got git compiling fine:
> - $ git bisect reset
> - $ git bisect start
> - $ git bisect good wine-1.1.4
> - $ git bisect bad wine-1.1.5
> - $ CC="ccache gcc" ./configure && make clean && make depend && make
> - $ ~/wine-git/wine your_program.exe
> - If good A) $ git bisect good
> If bad B) $ git bisect bad
>
> Repeat until you see something similar to:

By repeat, Austin meant go to step 5.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|major                       |normal

--- Comment #12 from Dmitry Timoshkov   2008-09-23 01:58:21 ---
Not major.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #13 from Ben Trusty   2008-09-23 17:22:30 ---
>

ok, so i did it again from scratch

almost done

<git bisect good>
d381c7cbf4d463189f479a0b3986f6767aeb8dba is first bad commit
commit d381c7cbf4d463189f479a0b3986f6767aeb8dba
Author: Alexandre Julliard 
Date:   Fri Sep 19 18:05:36 2008 +0200

Release 1.1.5.

:100644 100644 7a2c61113334222ab571b9cce093dd080fa825bf
982ac735f37215caf4e819c8afb6a8a2f0341585 M      ANNOUNCE
:100644 100644 cd289fc7fdb8187487084f08312e3042c1c7cb44
41c2c19075c1b563e1152512ed077b9895bd1f26 M      VERSION
:100755 100755 e898943c5e9326274904e3ecc42884b99456e288
ff2c11d3ae4e0afd80c9b62281c510521cef9e87 M      configure

<recompile again>
<run program - works>
<git bisect good>

d381c7cbf4d463189f479a0b3986f6767aeb8dba is first bad commit
commit d381c7cbf4d463189f479a0b3986f6767aeb8dba
Author: Alexandre Julliard 
Date:   Fri Sep 19 18:05:36 2008 +0200

Release 1.1.5.

:100644 100644 7a2c61113334222ab571b9cce093dd080fa825bf
982ac735f37215caf4e819c8afb6a8a2f0341585 M      ANNOUNCE
:100644 100644 cd289fc7fdb8187487084f08312e3042c1c7cb44
41c2c19075c1b563e1152512ed077b9895bd1f26 M      VERSION
:100755 100755 e898943c5e9326274904e3ecc42884b99456e288
ff2c11d3ae4e0afd80c9b62281c510521cef9e87 M      configure

<recompile again>
<run program - works>
<git bisect good>

d381c7cbf4d463189f479a0b3986f6767aeb8dba is first bad commit
commit d381c7cbf4d463189f479a0b3986f6767aeb8dba
Author: Alexandre Julliard 
Date:   Fri Sep 19 18:05:36 2008 +0200

Release 1.1.5.

:100644 100644 7a2c61113334222ab571b9cce093dd080fa825bf
982ac735f37215caf4e819c8afb6a8a2f0341585 M      ANNOUNCE
:100644 100644 cd289fc7fdb8187487084f08312e3042c1c7cb44
41c2c19075c1b563e1152512ed077b9895bd1f26 M      VERSION
:100755 100755 e898943c5e9326274904e3ecc42884b99456e288
ff2c11d3ae4e0afd80c9b62281c510521cef9e87 M      configure

so, same results 3 times after last 'remaining'
is there something else i need to do?
other than posting this out-put to the bugtracker?

(p.s. i did not copy/paste it multiple times, i did it AGAIN just to make sure
i did not do it too few times (as it appears that i did the first time.) i am
sure after doing it 3 times after it says (or as i assume) its done.)

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #14 from James Hawkins   2008-09-23 17:56:06 ---
Have you compiled 1.1.5 from git and made sure that the game is broken with
that build?  At any point during the regression test did you issue a git bisect
bad?  If it says 'first bad commit', then the test is done.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #15 from Ben Trusty   2008-09-23 18:45:10 ---
(In reply to comment #14)
> Have you compiled 1.1.5 from git and made sure that the game is broken with
> that build?  At any point during the regression test did you issue a git bisect
> bad?  If it says 'first bad commit', then the test is done.

is a compile from git 1.1.5 different from a 1.1.5 install?

the reason this whole thing started was b/c i did an update from 1.1.4 to 1.1.5
and then the game did not work. so i did the regression testing and started
this bug post.

if i have done
1. $ git bisect reset
2. $ git bisect start
3. $ git bisect good wine-1.1.4
4. $ git bisect bad wine-1.1.5
5. $ CC="ccache gcc" ./configure && make clean && make depend && make
6. $ ~/wine-git/wine your_program.exe
7. If good A) $ git bisect good
   If bad  B) $ git bisect bad

till it got to the point that it gave the same output (previous post), and
hasnt said "Bisecting: 0 revisions left to test after this" during the last
three times of doing processes 5-7.

what do i need to do now?
1. bisect reset (so i am using 1.1.5 code base)
2. CC="ccache gcc" ./configure && make clean && make depend && make
3. $ ~/wine-git/wine your_program.exe

and make sure that the game fails on 1.1.5 ?

then what?

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #16 from James Hawkins   2008-09-23 19:24:15 ---
(In reply to comment #15)
> is a compile from git 1.1.5 different from a 1.1.5 install?

Technically, yes, if by a 1.1.5 install, you mean an installed package from
your distribution.
> the reason this whole thing started was b/c i did an update from 1.1.4 to 1.1.5
> and then the game did not work. so i did the regression testing and started
> this bug post.

As I said before, a distro-packaged 1.1.5 can potentially be different from a
compiled git at 1.1.5.
> if i have done
> - $ git bisect reset
> - $ git bisect start
> - $ git bisect good wine-1.1.4
> - $ git bisect bad wine-1.1.5
> - $ CC="ccache gcc" ./configure && make clean && make depend && make
> - $ ~/wine-git/wine your_program.exe
> - If good A) $ git bisect good
> If bad  B) $ git bisect bad
>
> till it got to the point that it gave the same output (previous post), and
> hasnt said "Bisecting: 0 revisions left to test after this" during the last
> three times of doing processes 5-7.
>
> what do i need to do now?
> - bisect reset (so i am using 1.1.5 code base)
> - CC="ccache gcc" ./configure && make clean && make depend && make
> - $ ~/wine-git/wine your_program.exe

No, if your git is up to date, then you're past the 1.1.5 release.  To get back
to the 1.1.5 release,

$ echo d381c7cbf4d463189f479a0b3986f6767aeb8dba > .git/refs/heads/master
$ git checkout -f master
> and make sure that the game fails on 1.1.5 ?

Yes we need to know if it fails.  Please run it a couple times if it doesn't
fail, as the failure may be nondeterministic.  This may also be the reason why
the regression test failed, though I have my doubts, because you keep landing
on the same commit.  Please respond to the question, "Did you ever get a chance
to run a 'git bisect bad' command?"

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #17 from Austin English   2008-09-23 19:56:18 ---
Is there a demo of any of these games?

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #18 from Ben Trusty   2008-09-23 20:37:09 ---
(In reply to comment #17)
> Is there a demo of any of these games?

both command and conquer: tiberian dawn & command and conquer: red alert 1
are both freeware (they have been released since they are 12/13 years old by ea

tiberian dawn: http://www.fileplanet.com/55414/0/section/Command-&-Conquer
red alert: http://www.ea.com/redalert/news-detail.jsp?id=62

both games are 2 cds each.

to previous post:
i did

$ echo d381c7cbf4d463189f479a0b3986f6767aeb8dba > .git/refs/heads/master
$ git checkout -f master

recompiled
then ran the program 4-5 times
never failed
> Please respond to the question, "Did you ever get a chance
> to run a 'git bisect bad' command?"

no.

so the question is, how do i tell what is the difference between
git's 1.1.5
1.1.5~winehq0~ubuntu~8.04-0ubuntu1 release?

thanks for walking me through this

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

James Hawkins  changed:

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

--- Comment #19 from James Hawkins   2008-09-23 20:50:21 ---
(In reply to comment #18)
> to previous post:
> i did
>
> $ echo d381c7cbf4d463189f479a0b3986f6767aeb8dba > .git/refs/heads/master
> $ git checkout -f master
>
> recompiled
> then ran the program 4-5 times
> never failed
> > Please respond to the question, "Did you ever get a chance
> > to run a 'git bisect bad' command?"
>
> no.

Ok, so that's why the regression test didn't show a meaningful result.
> so the question is, how do i tell what is the difference between
> git's 1.1.5
> 1.1.5~winehq0~ubuntu~8.04-0ubuntu1 release?
>
> thanks for walking me through this

I don't know.  Unless I'm mistaken, Scott Ritchie packages wine for debian and
ubuntu.  I'm cc'ing him on this bug.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #20 from Scott Ritchie   2008-09-25 01:54:42 ---
So the 1.1.4 package worked for you and the 1.1.5 package didn't?

Have you changed any other packages in the interrim (eg upgrading to Intrepid)?

The package should be building the same way as before, unless the build
dependencies have changed substantially since 1.1.4 (which I don't think is the
case).

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #21 from Scott Ritchie   2008-09-25 01:55:47 ---
Also, to make sure the difference is in the packages and not something else
that might have changed, please uninstall 1.1.5 and reinstall 1.1.4 package to
make sure 1.1.4 still works.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Toni Spets  changed:

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

--- Comment #22 from Toni Spets   2008-12-08 12:20:11 ---
WORKSFORME in 1.1.9 if it speeds up processing this bug.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

--- Comment #23 from Austin English   2008-12-08 14:56:17 ---
(In reply to comment #22)
> WORKSFORME in 1.1.9 if it speeds up processing this bug.

If it works for you in 1.1.9, then the bug is (likely) fixed.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Xavier Vachon  changed:

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

--- Comment #24 from Xavier Vachon   2009-06-04 21:15:54 ---
With the schedtool workaround for multicore CPUs and latest git (1.1.22), I am
able to run both Command & Conquer Gold and Red Alert. I believe this is fixed.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Ken Sharp  changed:

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

--- Comment #25 from Ken Sharp   2009-06-05 17:00:07 ---
Reported fixed.

[Bug 15377] 1.1.5 breaks d3d game that worked in 1.1.4

http://bugs.winehq.org/show_bug.cgi?id=15377

Alexandre Julliard  changed:

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

--- Comment #26 from Alexandre Julliard   2009-06-19 11:00:52 ---
Closing bugs fixed in 1.1.24.