[Msys2-users] Current problems installing MinGW-w64/MSYS2 on the wine-staging platform

Discussion:

Alan W. Irwin

2017-06-16 21:18:41 UTC

The MinGW-w64/MSYS2 wiki at <https://github.com/msys2/msys2/wiki>
refers to <https://github.com/TeaCI/tea-ci/wiki/Msys2-on-Wine> for
instructions about how to install MinGW-w64/MSYS2 on the wine-staging
platform. Those instructions (updated a year ago by Qian Hong) fail
quite early in the process for me. I have attempted to contact Qian
Hong directly about that earlier this week, but he has not responded.
So although Qian Hong's wiki entry is obviously not part of the
MinGW-w64/MSYS2 effort, I thought I would at least detail what I tried
here in case someone else here knows how to get MinGW-w64/MSYS2
installed on the wine-staging platform.

My Linux platform is Debian Jessie so the first step was to install
wine-staging following the Debian Jessie instructions at
https://wine-staging.com/installation.html. That installed
winehq-staging 2.4 which is several months out of date, but I thought
it was likely new enough to attempt to proceed further since that
version was released after the above wiki entry was last updated in
July 2016.

I then set the following environment variables:

export WINEVERSION=2.4_jessie_wine-staging
export WINEPREFIX=/home/wine/wine_staging/.wine-$WINEVERSION
export WINEDEBUG='fixme-all'

and ran

winecfg

taking defaults for everything. I then ran these further commands:

wineserver -p
wget http://repo.msys2.org/distrib/x86_64/msys2-x86_64-20161025.exe
nice -19 wineconsole --backend=curses msys2-x86_64-20161025.exe

That GUI appeared to work fine for getting through the welcome page,
the prefix page (where I appended "_20170612" to the install prefix),
and the keyboard shortcuts page (where I used default), but the next
page where an attempt was made to actually install files, it got through
roughly 60 per cent of it and abruptly ended with what looked like a
bash.exe segfault and the following fatal
error message that came up on wineconsole

err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx
apartment not initialised

err:ole:CoCreateInstanceEx apartment
not initialised
wine: Unhandled page fault on read
access to 0x00000020 at address 0x7bc4b478 (thread 0046), starting
debugger...

Note the above installer is not what was recommended above in July
2016, but instead is the latest one you find if you start at
<https://github.com/msys2/msys2/wiki/MSYS2-installation> and follow
the 64-bit installation instructions.

I then repeated the attempt using the equivalent 32-bit installer,
msys2-i686-20161025.exe, but there were similar errors in that case as
well.

I must say I am quite surprised by these failures of the install GUI's
since my Wine experience with installers from several years ago for a
rather large variety of different software projects was that generally
install GUI's (whose simple job after all is normally to obtain files
from the web and install them) work well under Wine.

Is there something more I should be doing to get either of these
installers to work on wine-staging? If not, but someone here finds
these installers have no such issues under the latest wine-staging (as
opposed to the 2.4 version I am using), then can you tell me how do I
get access to that latest version?

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Norbert Pfeiler

2017-06-16 23:13:09 UTC

Permalink

Just a bit of field data:

Iâm seeing the same issues with the installer as well as running my windows
msys2 installation from Arch Linux with wine staging 2.10 (Unhandled page
fault on read).
The installer runs through smoothly with non-staging wine, but msys2 wonât
start showing stack overflow errors instead.

Best, Norbert.

Post by Alan W. Irwin
The MinGW-w64/MSYS2 wiki at <https://github.com/msys2/msys2/wiki>
refers to <https://github.com/TeaCI/tea-ci/wiki/Msys2-on-Wine> for
instructions about how to install MinGW-w64/MSYS2 on the wine-staging
platform. Those instructions (updated a year ago by Qian Hong) fail
quite early in the process for me. I have attempted to contact Qian
Hong directly about that earlier this week, but he has not responded.
So although Qian Hong's wiki entry is obviously not part of the
MinGW-w64/MSYS2 effort, I thought I would at least detail what I tried
here in case someone else here knows how to get MinGW-w64/MSYS2
installed on the wine-staging platform.
My Linux platform is Debian Jessie so the first step was to install
wine-staging following the Debian Jessie instructions at
https://wine-staging.com/installation.html. That installed
winehq-staging 2.4 which is several months out of date, but I thought
it was likely new enough to attempt to proceed further since that
version was released after the above wiki entry was last updated in
July 2016.
export WINEVERSION=2.4_jessie_wine-staging
export WINEPREFIX=/home/wine/wine_staging/.wine-$WINEVERSION
export WINEDEBUG='fixme-all'
and ran
winecfg
wineserver -p
wget http://repo.msys2.org/distrib/x86_64/msys2-x86_64-20161025.exe
nice -19 wineconsole --backend=curses msys2-x86_64-20161025.exe
That GUI appeared to work fine for getting through the welcome page,
the prefix page (where I appended "_20170612" to the install prefix),
and the keyboard shortcuts page (where I used default), but the next
page where an attempt was made to actually install files, it got through
roughly 60 per cent of it and abruptly ended with what looked like a
bash.exe segfault and the following fatal
error message that came up on wineconsole
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx
apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
wine: Unhandled page fault on read
access to 0x00000020 at address 0x7bc4b478 (thread 0046), starting
debugger...
Note the above installer is not what was recommended above in July
2016, but instead is the latest one you find if you start at
<https://github.com/msys2/msys2/wiki/MSYS2-installation> and follow
the 64-bit installation instructions.
I then repeated the attempt using the equivalent 32-bit installer,
msys2-i686-20161025.exe, but there were similar errors in that case as
well.
I must say I am quite surprised by these failures of the install GUI's
since my Wine experience with installers from several years ago for a
rather large variety of different software projects was that generally
install GUI's (whose simple job after all is normally to obtain files
from the web and install them) work well under Wine.
Is there something more I should be doing to get either of these
installers to work on wine-staging? If not, but someone here finds
these installers have no such issues under the latest wine-staging (as
opposed to the 2.4 version I am using), then can you tell me how do I
get access to that latest version?
Alan
__________________________
Alan W. Irwin
Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).
Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________
Linux-powered Science
__________________________
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Msys2-users mailing list
https://lists.sourceforge.net/lists/listinfo/msys2-users

Alan W. Irwin

2017-06-17 03:25:24 UTC

Permalink

I’m seeing the same issues with the installer as well as running my windows
msys2 installation from Arch Linux with wine staging 2.10 (Unhandled page
fault on read).

Hi Norbert:

Thanks for that confirmation of the issue.

Correction to my post below: I misread the version, (apt-cache lists
all versions fromn 2.10 back to 2.4, and I just read the last one), and it turns out I
am actually running wine-staging 2.9, and as soon as I update again,
I will be running wine-staging 2.10. However, from your result above, it
appears that update won't change anything. :-)

The installer runs through smoothly with non-staging wine

That is a really interesting result that suggests some
"fix" for wine-staging is interfering with the installer.

but msys2 won’t
start showing stack overflow errors instead.

Could you clarify what you mean by that last sentence?

Since MinGW-w64/MSYS2 has never worked properly on Wine without the
special patches Wine-staging has, I assume it is something like
"post-installer use of MinGW-w64/MSYS2 on non-staging Wine shows stack overflow
errors", but I want to make sure. But if you confirm that, my working
hypothesis (since apparently MinGW-w64/MSYS2 did work properly on
wine-staging in July 2016) is there are some patches on wine-staging
that are required, but another patch for wine-staging that is
interfering with the installer.

It appears that there is a package called wine on the wine-staging
Debian Jessie repository that conflicts with the winehq-staging
package I currently have installed. And the description of that
package refers to "standard" Wine. So I assume if I remove
winehq-staging and install wine, I should be able to confirm your
non-staging results.

Alan

__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Alan W. Irwin

2017-06-17 07:33:22 UTC

Permalink

Post by Alan W. Irwin
It appears that there is a package called wine on the wine-staging
Debian Jessie repository that conflicts with the winehq-staging
package I currently have installed. And the description of that
package refers to "standard" Wine. So I assume if I remove
winehq-staging and install wine, I should be able to confirm your
non-staging results.

Actually, that package is the ancient wine-1.x version from vanilla Debian
Jessie which should not be installed. The name of the package to be installed is actually called winehq-devel
which yields

***@raven> wine --version
wine-2.10

In contrast, that result when winehq-staging is installed is

wine-2.10 (Staging)

For the record, in the non-staging case I get the following results:

***@raven> nice -19 wine64 msys2-x86_64-20161025.exe
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
err:seh:setup_exception stack overflow 2240 bytes in thread 0043 eip
000000007bc949f2 esp 00000000ffe00d50 stack
0xffe00000-0xffe01000-0x100000000
***@raven> err:seh:setup_exception stack overflow 2240 bytes in
thread 004f eip 000000007bc949f2 esp 00000000ffe00d50 stack
0xffe00000-0xffe01000-0x100000000

where that stack overflow occurred after the file installation when the
GUI installer launches MSYS2.

In the staging case, I get the following results:

***@raven> nice -19 wine msys2-x86_64-20161025.exe
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
wine: Unhandled page fault on read access to 0x00000020 at address
0x7bc4b478 (thread 0045), starting debugger...

The installer GUI details showed the install failed (at the ~66 per
cent mark) while installing nansi.sys. That failure was accompanied
by Program Error GUI box with the message

The program bash.exe has encountered a serious error and needs to
close. We are sorry for the inconvenience.

The program error details page on GUI was empty.

So it appears we have at least one bug for vanilla wine-2.10 and one
bug for wine-staging-2.10.

For the vanilla case, the installer installs all files without issues
(other then the "apartment" warnings above), but has a stack overflow
when it attempts to actually start MSYS2 after the file installation.
(This result demonstrates that the well-known, long-standing issues
for MSYS2 on vanilla wine still exist for vanilla wine-2.10.)

For the wine-staging 2.10 case the installer fails even to install the
files. Presumably, that bug is caused by one of the recent wine-staging
patches, but if that issue can be figured out, there is likely a good
chance that MinGW-w64/MSYS2 would work on wine-staging since
<https://github.com/TeaCI/tea-ci/wiki/Msys2-on-Wine> implies that
procedure worked as of July 2016.

Alan

__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Alan W. Irwin

2017-06-17 08:23:55 UTC

Permalink

Post by Alan W. Irwin
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
err:ole:CoCreateInstanceEx apartment not initialised
wine: Unhandled page fault on read access to 0x00000020 at address
0x7bc4b478 (thread 0045), starting debugger...
The installer GUI details showed the install failed (at the ~66 per
cent mark) while installing nansi.sys. That failure was accompanied
by Program Error GUI box with the message
The program bash.exe has encountered a serious error and needs to
close. We are sorry for the inconvenience.
The program error details page on GUI was empty.

Assuming everything worked OK last July, there is a possibility that
this wine-staging bug is recently introduced. So I downgraded
the winehq-staging package and all its dependencies to the earliest
version available (2.4.0-3~jessie) in the Debian Jessie winehq
repository, but that made zero difference. That is, I replicated
the exact results above including the last file mentioned in the
installer GUI details being nansi.sys.

So it is obviously time for me to transcribe and reorganize these
results as an official wine-staging bug report.

Alan

__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Alan W. Irwin

2017-06-17 09:37:46 UTC

Permalink

Post by Alan W. Irwin
So it is obviously time for me to transcribe and reorganize these
results as an official wine-staging bug report.

See <https://bugs.winehq.org/show_bug.cgi?id=43193>. I would
appreciate it if those who can confirm this bug (like Norbert) add
their reports at that location since I imagine the more such
confirmations, the more likely the wine-staging developers will fix
this bug in a timely manner.

By the way, my motivation for attempting to install MSYS2 on the
wine-staging platform is I would like to build and test PLplot, the
timeephem projects ephcom and te_gen, and FreeEOS on that platform. My
prior experience with attempting to do the same thing on MinGW/MSYS
(several years ago when that platform was fairly reliable) was quite
successful although extremely slow because of the huge (typically 0.5
seconds) startup latency of executables being launched on MinGW/MSYS
(and presumably MinGW-w64/MSYS2 also).

My mental model of why that latency is so large (corresponding to
something like 100 million operations) is MinGW-w64/MSYS2 forking to
start applications is already quite complex (just as is also the case
for Cygwin). And each of those many Windows commands is translated by
Wine-staging (or Wine) into many Linux commands. So fundamentally
10000 ops * 10000 ops = 100 million ops to start up an application,
and there is no way around this latency issue.

That startup latency doesn't hurt gamers much since it only happens
when they start a game. That latency becomes a serious issue,
however, when building and testing software since most build systems
implement those tasks as a large number of interdependent small
applications that are run (each of which has that 0.5 second penalty
applied).

Despite the expected slowness due to startup latency, running MSYS2 on
wine-staging should be quite valuable to me since it allows me to
avoid paying a tax to a convicted monopolist that I dislike. However,
despite my personal distaste for Microsoft's attempts to extend its PC
monopoly, I do recognize many developers like to use Windows to
develop their own free software, and I don't mind at all giving them a
hand to do that because I feel free software is important regardless
of what operating system is running. Thus, once the above
wine-staging bug is fixed, and assuming the rest of MSYS2 works OK on
that platform as before, that should allow me to easily collaborate
with my software development colleagues that have access to MSYS2 on
the Microsoft version of the Windows platform.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________