[Bug 11537] New: Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

Summary: Sid Meier's Pirates! install fails with 1603 (failing to
                    request disk 2)
           Product: Wine
           Version: 0.9.55.
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: msi
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]

Sid Meier's pirates install fails.  The installation returns a 1603 error,
while the installer reports that it is copying 2K Games.url.

The console output shows:

err:msi:ready_media Cabinet not found: L"E:\mnt\pirates1\Language.cab"
err:msi:ACTION_InstallFiles Failed to ready media
err:msi:ITERATE_Actions Execution halted, action L"InstallFinalize" returned
1603
err:ole:ClientRpcChannelBuffer_SendReceive called from wrong apartment, should
have been 0x1500000031
err:ole:xCall RpcChannelBuffer SendReceive failed, 8001010e

Language.cab is on Disk 2, so the installer should have requested that I switch
disks.  I can personally attest this is a regression since 0.9.29, and AppDB
suggests it is a regression is 0.9.52.  (There were no tests on 53 or 54)

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #1 from Austin English   2008-02-11 14:32:14 ---
Can you please run a regression test:

http://wiki.winehq.org/RegressionTesting

Also, please attach full terminal output.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #2 from Stephen E. Baker   2008-02-12 09:52:44 ---
Performing a regression test between 0.9.45 (known good) and 0.9.50 (known bad)
I ran into another regression at wine-0.9.49-196-ga6dd479 which prevents
testing this issue.  Is there some way to get around this regression without
messing up the binary search?

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #3 from Stephen E. Baker   2008-02-12 10:44:00 ---
(In reply to comment #2)
> Performing a regression test between 0.9.45 (known good) and 0.9.50 (known bad)
> I ran into another regression at wine-0.9.49-196-ga6dd479 which prevents
> testing this issue.  Is there some way to get around this regression without
> messing up the binary search?

It appears that the installer is largely unstable between 0.9.49 where it
worked, and 0.9.50 where this bug first occurred.

Here are my results:

wine-0.9.49 good
wine-0.9.49-196-ga6dd479 untestable (cannot run install shield)
wine-0.9.49-235-ga0c3816 untestable (fails to copy first file with 1603)
wine-0.9.50 bad

What should I do next to help out?

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #4 from Stephen E. Baker   2008-02-12 11:07:00 ---
Created an attachment (id=10727)
 --> (http://bugs.winehq.org/attachment.cgi?id=10727)
wine-0.9.50 Console Output

Console output from 0.9.50 (the first release version with this bug)

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #5 from Austin English   2008-02-12 11:39:46 ---
(In reply to comment #2)
> Performing a regression test between 0.9.45 (known good) and 0.9.50 (known bad)
> I ran into another regression at wine-0.9.49-196-ga6dd479 which prevents
> testing this issue.  Is there some way to get around this regression without
> messing up the binary search?

Can you describe the regression better? You could try resetting git to
0.9.46/7, see if they're good/bad, then use that as the known good/bad in a
separate regression test. Without knowing what's wrong, it's hard to say.
> It appears that the installer is largely unstable between 0.9.49 where it
> worked, and 0.9.50 where this bug first occurred.
>
> Here are my results:
>
> wine-0.9.49 good
> wine-0.9.49-196-ga6dd479 untestable (cannot run install shield)
> wine-0.9.49-235-ga0c3816 untestable (fails to copy first file with 1603)
> wine-0.9.50 bad
>
> What should I do next to help out?

If this is the same regression, then you can try 'random' patches in between,
or you could find the patch(es) that fix install shield using reverse
regression testing:
http://wiki.winehq.org/ReverseRegressionTesting

In short, describe your problem(s) more, but it doesn't look like an easy
solution.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #6 from Stephen E. Baker   2008-02-12 12:02:47 ---
(In reply to comment #5)
> (In reply to comment #2)
> > Performing a regression test between 0.9.45 (known good) and 0.9.50 (known bad)
> > I ran into another regression at wine-0.9.49-196-ga6dd479 which prevents
> > testing this issue.  Is there some way to get around this regression without
> > messing up the binary search?
>
> Can you describe the regression better? You could try resetting git to
> 0.9.46/7, see if they're good/bad, then use that as the known good/bad in a
> separate regression test. Without knowing what's wrong, it's hard to say.

Sorry, I shouldn't have replied to this post.  I reset git to 0.9.49 and it was
good, as have been all versions tested before that point so I have no reason to
suspect any problems with 46/47 etc.
> > It appears that the installer is largely unstable between 0.9.49 where it
> > worked, and 0.9.50 where this bug first occurred.
> >
> > Here are my results:
> >
> > wine-0.9.49 good
> > wine-0.9.49-196-ga6dd479 untestable (cannot run install shield)
> > wine-0.9.49-235-ga0c3816 untestable (fails to copy first file with 1603)
> > wine-0.9.50 bad
> >
> > What should I do next to help out?
>
> If this is the same regression, then you can try 'random' patches in between,
> or you could find the patch(es) that fix install shield using reverse
> regression testing:
> http://wiki.winehq.org/ReverseRegressionTesting
>
> In short, describe your problem(s) more, but it doesn't look like an easy
> solution.

Pirates! uses two cds, and the installer should prompt the user to switch cds
about half way through the install to copy the rest of the programs files from.
 (The files are stored for the most part in a series of .cab files.)

In versions of wine up to 0.9.49, this worked properly.  The installer would
prompt that the user switch cds, and when they clicked ok it would resume the
install from the second cd.
> From 0.9.50 onwards, the installer no longer prompts the user for the second

cd.  It instead continues to attempt to use a file which is located on the
second cd, and not finding it, displays a 1603 error and aborts.

James Hawkins made most of the changes during this period and a lot of them
seem to be related to cabinet files, so he is likely the best person to ask.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #7 from James Hawkins   2008-02-12 12:11:02 ---
As there were lots of changes made, we still need the results of the regression
test between 0.9.49 and 0.9.50.  As Austin said, you can use reverse regression
testing to find the patch that fixed the installshield problem and use that
patch when doing the actual regression test.  Same for any other problems
encountered during a regression test.  In the meantime, please attach a
compressed +msi,+msidb log and the .msi file (compressed).

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #8 from Stephen E. Baker   2008-02-12 13:00:06 ---
Created an attachment (id=10731)
 --> (http://bugs.winehq.org/attachment.cgi?id=10731)
WINEDEBUG=warn+msi,warn+msidb ./wine /mnt/pirates1/setup.exe

msi debugging output from 0.9.50 (compressed with gzip)

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #9 from Stephen E. Baker   2008-02-12 13:00:41 ---
Created an attachment (id=10732)
 --> (http://bugs.winehq.org/attachment.cgi?id=10732)
msi file for Sid Meier's Pirates!

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #10 from Stephen E. Baker   2008-02-12 15:28:35 ---
(In reply to comment #7)
> As there were lots of changes made, we still need the results of the regression
> test between 0.9.49 and 0.9.50.  As Austin said, you can use reverse regression
> testing to find the patch that fixed the installshield problem and use that
> patch when doing the actual regression test.  Same for any other problems
> encountered during a regression test.  In the meantime, please attach a
> compressed +msi,+msidb log and the .msi file (compressed).

wine-0.9.49-264-g075e84b fixes the bug where the installer would give a 1603
error and abort copying the first file.  This same revision exhibits my bug,
where it gives a 1603 and aborts when it should ask for the second disk.

I have a feeling that this bug was produced during a time when it could not be
tested or along with another bug that was fixed later.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #11 from James Hawkins   2008-02-12 15:39:15 ---
Stephen, you know the patch that fixes the installshield problem, so when doing
the regression test and you experience the installshield bug, apply the patch
from said commit, rebuild and try again.  This way you *can* do the regression
test.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #12 from Stephen E. Baker   2008-02-12 21:55:07 ---
I will get back to this tomorrow, using wine-0.9.49-216-g78eead9 to create a
patch against earlier revisions suffering from *1.  If anyone is working on
this in the mean time hopefully these git regression test results can help. (I
have left only the noteworthy results)

wine-0.9.49 *good
wine-0.9.49-215-g05a9678 untestable (*1)
wine-0.9.49-216-g78eead9+264patch *bad
wine-0.9.49-264-g075e84b *bad (patch from this = 264patch, needed for earlier
revisions)
wine-0.9.50 *bad

*1 - 1607: Unable to install InstallShield Scripting Runtime regression
*bad - suffers 'not prompting for second disk regression'
*good - no regression

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #13 from Stephen E. Baker   2008-02-13 14:46:43 ---
Ok, the significant portion of my regression results

wine-0.9.49 *good
wine-0.9.49-159-ga97d655 *good
wine-0.9.49-160-gbb747e4 (*1) + 162patch (*1) + 216patch (*2) + 264patch
(*bad2)
wine-0.9.49-161-gbb747e4 (*1) + 216patch (*3) untestable
wine-0.9.49-162-gb1507ae (*1) + 216patch (*2) + 264patch (*bad2)
wine-0.9.49-216-g78eead9 (*4) + 264patch (*bad)
wine-0.9.49-217-g3c8663b (*4) + 264patch (*bad)
wine-0.9.49-264-g075e84b *bad
wine-0.9.50 *bad

*1 - 1607: Unable to install InstallShield Scripting Runtime regression
*2 - 1627: ERROR_FUNCTION_FAILED
*3 - Patch fails
*4 - 1603 error when starting to copy files
*bad - suffers 'not prompting for second disk regression' with 1603 error
*bad2 - same as bad, but error message is 1627: ERROR_FUNCTION FAILED
*good - no regression

So as you can see, wine-0.9.49-160 patched with 162, 216, and 264 is the first
version to exhibit the bug.  I haven't traced down where the slight differences
in the error message came from, but it's likely the result of a minor change in
dlls/msi/files.c unrelated to my bug.  Hopefully with this information you will
be able to find how it use to detect that the second disk was needed and why it
isn't now.

Note: Prior to the regression the switch disks dialog had an OK and Cancel
button, but clicking the Cnacel button would just reprompt you.  Perhaps this
could be fixed at the same time.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #14 from James Hawkins   2008-02-13 16:35:30 ---
I need the actual commit id.  wine-0.9.49-160 doesn't mean anything to me.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #15 from Stephen E. Baker   2008-02-13 20:45:24 ---
(In reply to comment #14)
> I need the actual commit id.  wine-0.9.49-160 doesn't mean anything to me.

They were all in that last comment: wine-0.9.49-160-gbb747e4

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #16 from James Hawkins   2008-02-13 21:40:49 ---
(In reply to comment #15)
> They were all in that last comment: wine-0.9.49-160-gbb747e4

That's not a commit id.  This is an example of a commit id:

commit 14f7a59270579d57244e7ce3527f157233f0bc3b

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #17 from Lei Zhang   2008-02-13 23:01:39 ---
(In reply to comment #16)
> (In reply to comment #15)
> > They were all in that last comment: wine-0.9.49-160-gbb747e4
>
> That's not a commit id.  This is an example of a commit id:
>
> commit 14f7a59270579d57244e7ce3527f157233f0bc3b

the commit id is probably bb747e4fbe6619fa0e862102cc26456118355c67

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #18 from Stephen E. Baker   2008-02-14 05:43:39 ---
(In reply to comment #17)
> (In reply to comment #16)
> > (In reply to comment #15)
> > > They were all in that last comment: wine-0.9.49-160-gbb747e4
> >
> > That's not a commit id.  This is an example of a commit id:
> >
> > commit 14f7a59270579d57244e7ce3527f157233f0bc3b
>
> the commit id is probably bb747e4fbe6619fa0e862102cc26456118355c67

That it correct. It's easy to get this value from the numbers I gave, as they
are all git needs to move to a particular commit.

git reset --hard wine-0.9.49-xxx-xxxxxxx
git show | grep commit

anyway the commits which might have caused the bug are: 
wine-0.9.49-160-gbb747e4: bb747e4fbe6619fa0e862102cc26456118355c67
wine-0.9.49-162-gb1507ae: b1507aee989b168e8c08d0202ac60e467db25871
wine-0.9.49-216-g78eead9: 78eead93fd874634fdffd7c93e015bca4da0d755
wine-0.9.49-264-g075e84b: 075e84bd90bc5dae9cb050d110281ce8ea1b7256

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #19 from James Hawkins   2008-02-14 12:10:25 ---
I can't afford time on this bug unless I have *the* commit that caused the bug.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #20 from Stephen E. Baker   2008-02-14 13:11:24 ---
(In reply to comment #19)
> I can't afford time on this bug unless I have *the* commit that caused the bug.

160 is the commit with the label 'Simplify ready-media' which touches the: "if
(type == DRIVE_CDROM || type == DRIVE_REMOVABLE) lines" and moves the
msi_change_media call.  I only listed the others because this commit broke
other things and those other commits were needed to get them working again, so
they should be considered as a whole.

Assuming 160 is the cause of the problem, the only method you have to look at
is ready_media; however I don't really know how msi/wine works.  Does
ready_media get called for each file, or just once per media?  I am a
developer, I've just never worked with WINE, so with a little bit of guidance I
may be able to write the patch if you don't have time.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

Austin English  changed:

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

--- Comment #21 from Austin English   2008-02-14 16:11:47 ---
(In reply to comment #20)
> (In reply to comment #19)
> > I can't afford time on this bug unless I have *the* commit that caused the bug.
>
> 160 is the commit with the label 'Simplify ready-media' which touches the: "if
> (type == DRIVE_CDROM || type == DRIVE_REMOVABLE) lines" and moves the
> msi_change_media call.  I only listed the others because this commit broke
> other things and those other commits were needed to get them working again, so
> they should be considered as a whole.

Stephen, in the future, please post the commit id. It is much easier to
find/fix a problem with this information.

James, here's the commit:

bb747e4fbe6619fa0e862102cc26456118355c67 - msi: Simplify ready_media.

http://source.winehq.org/git/wine.git/?a=commit;h=bb747e4fbe6619fa0e862102cc...

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #22 from Stephen E. Baker   2008-02-15 06:52:29 ---
> James, here's the commit:
>
> bb747e4fbe6619fa0e862102cc26456118355c67 - msi: Simplify ready_media.
>
> http://source.winehq.org/git/wine.git/?a=commit;h=bb747e4fbe6619fa0e862102cc...

I've looked over the diff in detail and it appears that there is one condition
where the results would be different:

if (file->IsCompressed &&
       GetFileAttributesW(mi->source) == INVALID_FILE_ATTRIBUTES) &&
       (!(package->BaseURL && UrlIsW(package->BaseURL, URLIS_URL))) && 
       ((!(mi->volume_label && mi->disk_id > 1))||
         source_matches_volume(mi, source_dir)||
         (!(type==DRIVE_CDROM || type==DRIVE_REMOVEABLE)));

In this case, the pre-commit would default to requesting:
msi_change_media(package, mi);

and the post-commit would:
ERR("Cabinet not found: %s\n", debugstr_w(mi->cabinet));
return ERROR_INSTALL_FAILURE;

I haven't figured out how to use the winedbg yet well enough to check these
conditions, perhaps someone else could?  In the mean time I assume simply
replacing the error with msi_change_media would be too much of a hack.

This is in: 
static UINT ready_media(MSIPACKAGE *package, MSIFILE *file, struct media_info
*mi) in dlls/msi/files.c

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

Stephen E. Baker  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |Installer

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #23 from James Hawkins   2008-02-16 18:15:05 ---
Do you have the CD device properly set up in winecfg?
Are you using physical media or ISOs?
If you're using physical media, how are you ejecting the first CD?
If you're using ISOs, do you own the loopback device?

You should know that I just installed this app and it worked fine.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #24 from Stephen E. Baker   2008-02-17 09:47:58 ---
(In reply to comment #23)
> Do you have the CD device properly set up in winecfg?
> Are you using physical media or ISOs?
> If you're using physical media, how are you ejecting the first CD?
> If you're using ISOs, do you own the loopback device?
>
> You should know that I just installed this app and it worked fine.

I first ran into this problem on the physical media.  I have verified that H:
in winecfg is path: /mnt/cdrom, type: CD-ROM, and when Pirates disk 1 is in
uses label: Disc 1, Serial: D11D6EB3.

For regression testing purposes I dd'd the iso and created a loopback device. 
It is owned by my user, and is also set up properly in winecfg.

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #25 from James Hawkins   2008-02-17 11:51:45 ---
How are you running the installer (complete command line, current working
directory)?

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

--- Comment #26 from James Hawkins   2008-02-17 13:17:57 ---
You say the CD device is set up as H: in winecfg, yet your log says it's
looking for the cab at "E:\mnt\pirates1\Language.cab".  That means there's
something invalid with your configuration or you're not running the executable
correctly (PWD).

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

Stephen E. Baker  changed:

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

--- Comment #27 from Stephen E. Baker   2008-02-17 15:54:51 ---
Installing from the cdrom (./wine /mnt/cdrom/setup.exe) mentions looking for
H:\Language.cab and still gives the 1603.  However, when I moved my .wine out
of the way to use a fresh configuration the install worked fine.  I'm not sure
what's going on in the old configuration, but I assume the reason it only
appears broken after 0.9.49 is the tighter (but valid) conditions on swapping
media.  I'll mark this as invalid. (sorry about wasting everyone's time)

[Bug 11537] Sid Meier's Pirates! install fails with 1603 ( failing to request disk 2)

http://bugs.winehq.org/show_bug.cgi?id=11537

James Hawkins  changed:

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

--- Comment #28 from James Hawkins   2008-02-17 17:43:43 ---
Closing invalid.