VC++ nmake not working with Wine 20030115 when it worked with 200208xx

[J. Grant]

Hi,

Calling nmake from /wcmd/ used to work, now it does not.  I can't see
why not, here is the debugger ouput if anyone knows whats wrong I'd be
happy to hear!

Breakpoint 1 at 0x40009bd0 (_end+0x37fae414)

*** Invalid address 0x40011f80 (_end+0x37fb67c4)

<snip>
Unhandled exception: unimplemented function msvcrt.dll.__p__pgmptr
called in 32-
bit code (0x40733056).

In 32-bit mode.

0x40733056 (__wine_unimplemented+0x56 [msvcrt.spec.c:47] in
msvcrt.dll.so): jmp
0x40733050 (__wine_unimplemented+0x50 [msvcrt.spec.c:47] in
msvcrt.dll.so)
47        for (;;) RtlRaiseException( &rec );

Wine-dbg>

There are a few unknown functions with "__p__pgmptr"

void __wine_stub_msvcrt_dll_122(void) {
__wine_unimplemented("__p__pgmptr"); }

I'm not sure why this problem occures now, when it worked before. 
Comments welcome

Regards

JG

[Tony Lambregts]

J. Grant wrote:
> Hi,
>
> Calling nmake from /wcmd/ used to work, now it does not.  I can't see
> why not, here is the debugger ouput if anyone knows whats wrong I'd be
> happy to hear!

This is called a regression. I see you are not on CVS but if you switch 
to it, you could find the the patch that broke the app. The link on how 
to do this is here.

http://www.winehq.org/docs/wine-devel/cvs-regression.shtml

If you have problems with this or have suggestions on how to improve the 
docu please let me know.

[J. Grant]

Hi Tony,
> This is called a regression. I see you are not on CVS but if you switch 
> to it, you could find the the patch that broke the app. The link on how 
> to do this is here.
>
> http://www.winehq.org/docs/wine-devel/cvs-regression.shtml

OK, thank you for the infomation, very well written document!

Unfortunatly I am on a momdem, which restricts my abilities here.  I
will look for a workaround to find the regression patch.
> If you have problems with this or have suggestions on how to improve the 
> docu please let me know.

Well, could "Cvs" be written "CVS" ? The technical infomation seems top
notch.

Regards

JG

[Tony Lambregts]

J. Grant wrote:
> Hi Tony,
> > This is called a regression. I see you are not on CVS but if you 
> > switch to it, you could find the the patch that broke the app. The 
> > link on how to do this is here.
> >
> > http://www.winehq.org/docs/wine-devel/cvs-regression.shtml
>
> OK, thank you for the infomation, very well written document!
>
> Unfortunatly I am on a momdem, which restricts my abilities here.  I
> will look for a workaround to find the regression patch.

I did regression testing via modem in the past.  "cvs -z 3" compresses 
the stream so that download times are minimized.

The number of downloads required to find the day.

Days No of itinerations 
1    1
2    2
4    3
8    4
16   5
32   6
64   7
128  8
256  9
512  10

Assumming that there are no masking regressions to complicate the 
process you should be able to find the day after nine tests. After you 
have the day the number of substitue the number of patches for days to 
fine the itinerations required.  Most days  it would be 6 or less.

The biggest problem with regression testing for me was not the download 
time but the compile time. which is dependent on the computers speed and 
amount of RAM.  On my old P300 with 96mb  it can take up to three hours.

I obviously cant force you to do this but regression testing is the most 
efficient way to fix these problems. As a benifit of setting this up you 
get to be bleading edge up to date as possible, thus making it possible 
to find regressions very quickly. I actually have three trees one for 
development, one for regression testing and one "clean" up to date CVS.

One last thing. If you do decide to do this please first check that it 
is not fixed in an up-to-date  CVS (it happens!)
> > If you have problems with this or have suggestions on how to improve 
> > the docu please let me know.
>
> Well, could "Cvs" be written "CVS" ? The technical infomation seems top
> notch.

Can do. I see a few other minor thing I want to clean up while i'm at it.

[J. Grant]

Hi Tony,
> I did regression testing via modem in the past.  "cvs -z 3" compresses 
> the stream so that download times are minimized. 
> The number of downloads required to find the day.

OK, thaks for this useful infomation! I have a 1GHz athlon so it should 
not be too painful. I will give it a try over the next day or so.
> One last thing. If you do decide to do this please first check that it 
> is not fixed in an up-to-date  CVS (it happens!)

Yes, I was planning on checking that, and also Wine 20020804 incase it 
was my origional build that was at fault.

Regards

JG

[J. Grant]

Hi Tony,
> I did regression testing via modem in the past.  "cvs -z 3" compresses

I have been looking for the best way to tackle this.

http://www.winehq.org/docs/wine-devel/cvs-regression.shtml

On this page cvs-dirs-2000-05-20.tar.gz is used, but I can not find a 
file starting with this filename.

http://winehq.com/development/   mentions full-cvs-date.tar.gz as a full 
copy of the repository. available at ftp://ftp.winehq.com/pub/wine/  If 
these are the correct filenames instead of cvs-dirs-* (there were some 
files called wine-cvsdirs-* but they were only 40KB) could this be 
amended please?

So, I am going to download this, as it looks big enough! 50MB! Hope its 
the right archive?  ftp://ftp.winehq.com/pub/wine/full-cvs-2003-02-12.tar.gz

Would it be possible to add this one line command to get the wine src to 
the development page ?(http://winehq.com/development/) I think its 
simpler to read, and might attract more people to wine..

cvs -z 3 -d :pserver:[e-mail]:/home/wine co wine

cvs -z 3 -d :pserver:[e-mail]:/home/wine co wine

Regards

JG

[Vincent Béron]

J. Grant a écrit:
> Hi Tony,
> > I did regression testing via modem in the past.  "cvs -z 3" compresses
>
> I have been looking for the best way to tackle this.
>
> http://www.winehq.org/docs/wine-devel/cvs-regression.shtml
>
> On this page cvs-dirs-2000-05-20.tar.gz is used, but I can not find a 
> file starting with this filename.
>
> http://winehq.com/development/   mentions full-cvs-date.tar.gz as a full 
> copy of the repository. available at ftp://ftp.winehq.com/pub/wine/  If 
> these are the correct filenames instead of cvs-dirs-* (there were some 
> files called wine-cvsdirs-* but they were only 40KB) could this be 
> amended please?
>
> So, I am going to download this, as it looks big enough! 50MB! Hope its 
> the right archive?  
> ftp://ftp.winehq.com/pub/wine/full-cvs-2003-02-12.tar.gz

This is most probably not what you want. If you grab this file, you'll 
have every (as in, since 1998 or so) revision of every file in the Wine 
CVS repository. Although with that, you won't need a net connection to 
do regressions before the day you downloaded it...

The wine-cvsdirs-* is the archive you want. It only contains the 
directories CVS uses to know which revision of each file is the current 
one in your tree. In this case, it's the one in the corresponding 
snapshot. Once installed the wine-cvsdirs-*, you'll be able to get a 
copy of the tree as of different times in the past, and locate when the 
problem you're looking for arose.

Vincent

[J. Grant]

Salut Vincent,

Thanks for your reply.
> This is most probably not what you want. If you grab this file, you'll 
> have every (as in, since 1998 or so) revision of every file in the Wine 
> CVS repository. Although with that, you won't need a net connection to 
> do regressions before the day you downloaded it...

Yes, this looked simplest, otherwise I'm on modem waiting while I do 
stuff, i would rather download at night and have it ready to work in the 
day.  This is what the cvs-regression.shtml article suggested first it 
seemed.
> The wine-cvsdirs-* is the archive you want. It only contains the 
> directories CVS uses to know which revision of each file is the current 
> one in your tree. In this case, it's the one in the corresponding 
> snapshot. Once installed the wine-cvsdirs-*, you'll be able to get a 
> copy of the tree as of different times in the past, and locate when the 
> problem you're looking for arose.

OK, thanks, I will bare that in mind when working on this.

Regards

JG

[Tony Lambregts]

J. Grant wrote:
> Hi Tony,
> > I did regression testing via modem in the past. "cvs -z 3" compresses
>
> I have been looking for the best way to tackle this.
>
> http://www.winehq.org/docs/wine-devel/cvs-regression.shtml
>
> On this page cvs-dirs-2000-05-20.tar.gz is used, but I can not find a 
> file starting with this filename.
>
> http://winehq.com/development/ mentions full-cvs-date.tar.gz as a full 
> copy of the repository. available at ftp://ftp.winehq.com/pub/wine/ If 
> these are the correct filenames instead of cvs-dirs-* (there were some 
> files called wine-cvsdirs-* but they were only 40KB) could this be 
> amended please?
>
> So, I am going to download this, as it looks big enough! 50MB! Hope 
> its the right archive? 
> ftp://ftp.winehq.com/pub/wine/full-cvs-2003-02-12.tar.gz

What Vincent said about this is correct.
> Would it be possible to add this one line command to get the wine src 
> to the development page ?(http://winehq.com/development/) I think its 
> simpler to read, and might attract more people to wine..
>
> cvs -z 3 -d :pserver:[e-mail]:/home/wine co wine
>
> cvs -z 3 -d :pserver:[e-mail]:/home/wine co wine

Well if you are on a modem this might take a while!

[Dan Kegel]

Tony Lambregts wrote:
> > > http://www.winehq.org/docs/wine-devel/cvs-regression.shtml
> >
> > Unfortunatly I am on a momdem, which restricts my abilities here.  I
> > will look for a workaround to find the regression patch.
>
> I did regression testing via modem in the past.  "cvs -z 3" compresses 
> the stream so that download times are minimized.

I found it much faster to simply download the raw CVS archives;
do that once, and the cvs gets from the local archive run a
zillion times faster.  I think it's a win if you have to do
more than four or five downloads.
- Dan

[Eric Pouech]

J. Grant wrote:
> Hi,
>
> Calling nmake from /wcmd/ used to work, now it does not.  I can't see
> why not, here is the debugger ouput if anyone knows whats wrong I'd be
> happy to hear!

did you upgrade VC in the meantime ?
were you using native or builtin msvcrt ?

A+

[J. Grant]

Salut,

Bien recu votre email.
> did you upgrade VC in the meantime ?

Non, c'est tout comme avant.
> were you using native or builtin msvcrt ?

Well, "config" [DllOverrides]  has this line below, and I placed the
msvcrt.dll in the mnake.exe dir, so I believe it could have been using 
it before.

config line:
"msvcrt"       = "native, builtin"

msvcrt.dll is still in the bin dir, so when i run nmake from wcmd or on 
a bash shell wine --.namke.exe it still has the same problem.  Just in 
case I put some common dlls that I used to get other programs running 
(in the specific dir again) and called nmake again.  Wine debugger 
started agian with the same issue as before.

from bash I ran:
wine -dll msvcrt,comctl32,comdlg32,mspdb60,commdlg=n 
d:\vs\vc98\bin\nmake.exe

Regards

JG

[J. Grant]

OK,

Eric Pouech wrote:
<snip>
> were you using native or builtin msvcrt ?

I installed Wine 20020804 again, and it is working again, it was using 
msvcrt.dll that I had placed in the bin dir.  I get the following error 
without it, and the command returns without displaying anything.

fixme:msvcrt:_XcptFilter (0,0x40626b58)semi-stub
wine client error:0x80bd8b0: partial write 64

Regards

JG

[Tony Lambregts]

J. Grant wrote:
> OK,
>
> Eric Pouech wrote:
>
> <snip>
> > were you using native or builtin msvcrt ?
>
> I installed Wine 20020804 again, and it is working again, it was using 
> msvcrt.dll that I had placed in the bin dir. I get the following error 
> without it, and the command returns without displaying anything.
>
> fixme:msvcrt:_XcptFilter (0,0x40626b58)semi-stub
> wine client error:0x80bd8b0: partial write 64

Sounds like this bug then.

http://bugs.winehq.com/show_bug.cgi?id=1257

it is fixed by this patch (not yet in CVS)

http://www.winehq.com/hypermail/wine-patches/2003/02/0098.html

[J. Grant]

Hi Tony,

Thanks for the info.
> > fixme:msvcrt:_XcptFilter (0,0x40626b58)semi-stub
> > wine client error:0x80bd8b0: partial write 64
>
> Sounds like this bug then.
>
> http://bugs.winehq.com/show_bug.cgi?id=1257
>
> it is fixed by this patch (not yet in CVS)
>
> http://www.winehq.com/hypermail/wine-patches/2003/02/0098.html

Just read that, so it appears that a work around is to copy the dll to 
windows/system dir?

Well, I just did that with msvcrt.dll and unfortunatly it still crashes 
with this

wine ./nmake.exe   or
wine -dll msvcrt=n ./nmake.exe

This leads me to think that that this is not the same bug

with the newer wine build /nmake.exe/ and /cl.exe/ do not work.  With 
the following error:
fixme:msvcrt:_XcptFilter (-2147483392,0x40702720)semi-stub

Also /cmd.exe/ does not work: (is this related?)

$ wine ./cmd.exe
fixme:console:SetConsoleCtrlHandler (0x4ad09d41,1) - no error checking 
or testing yet
A subdirectory or file .exe already exists.

/link.exe/ does run, and so do /winemine/ and /notepad/, so I think the 
install must be ok.

I could not find any bugs relating to XcptFilter, are you aware of any 
issues that could resolve this or workaround it please?

Regards

JG

[Tony Lambregts]

J. Grant wrote:
> Hi Tony,
>
> Thanks for the info.
> > > fixme:msvcrt:_XcptFilter (0,0x40626b58)semi-stub
> > > wine client error:0x80bd8b0: partial write 64
> >
> > Sounds like this bug then.
> >
> > http://bugs.winehq.com/show_bug.cgi?id=1257
> >
> > it is fixed by this patch (not yet in CVS)
> >
> > http://www.winehq.com/hypermail/wine-patches/2003/02/0098.html
>
> Just read that, so it appears that a work around is to copy the dll to 
> windows/system dir?

No. move,  not copy, as long as the dll is in the programs directory 
wine will use the built-in (wines) version of the dll (bad bug! fifty 
pushups) This bug was introduced 2002/09/24 19:16:52 so it would affect 
your latter (20030115) tarball.
> Well, I just did that with msvcrt.dll and unfortunatly it still 
> crashes with this
>
> wine ./nmake.exe   or
> wine -dll msvcrt=n ./nmake.exe
>
> This leads me to think that that this is not the same bug
>
> with the newer wine build /nmake.exe/ and /cl.exe/ do not work.  With 
> the following error:
> fixme:msvcrt:_XcptFilter (-2147483392,0x40702720)semi-stub

this is the built-in (wine) version of msvcrt that is reporting this. 
 Your program is expecting its own version of msvcrt.

[J. Grant]

Hi
> No. move,  not copy, as long as the dll is in the programs directory 
> wine will use the built-in (wines) version of the dll (bad bug! fifty 
> pushups) This bug was introduced 2002/09/24 19:16:52 so it would affect 
> your latter (20030115) tarball.

OK, it is now running fine again with this work around.  Is there any 
news on when this will be fixed in the CVS version of Wine?
> > with the newer wine build /nmake.exe/ and /cl.exe/ do not work.  With 
> > the following error:
> > fixme:msvcrt:_XcptFilter (-2147483392,0x40702720)semi-stub
>
> this is the built-in (wine) version of msvcrt that is reporting this. 
> Your program is expecting its own version of msvcrt.

OK, thanks

JG