Re: Failure with fork() (Cygwin on Wine install now sort of works)

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Alan W. Irwin]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1994 bytes --]

Addendum.  I just noticed the request for cygcheck results at
http://cygwin.com/problems.html.  Accordingly,

I put /z/home/wine/newstart/cygwin/bin at the top of my
MSYS (on Wine) PATH, then executed

bash.exe-3.1$ cygcheck.exe -s -v -r >cygcheck.out

That generated a response to the terminal which was

cygcheck: Wrong architecture. Only ix86 executables supported.

For the record, I am using 32-bit Cygwin on 32-bit Wine on 64-bit
(x86_64) hardware.  My Linux platform is Debian wheezy with mostly
64-bit libraries installed, but 32-bit Wine was built specifically
with some extra 32-bit libraries I specifically installed for that
purpose.  So I don't know the source of that interesting message.
Note, that message occurs for the part of the cygchek results that
depend on cygwin1.dll.  If I fail to put cygwin/bin on my PATH,
then cygcheck complains that cygwin1.dll is not available but also
does not generate the above message.

I attach the cygcheck.out file generated as above in case somebody can
spot anything wrong in that report with my initial Cygwin on Wine
install. But superficially, it looks OK to me so it is possible that
my Cygwin on Wine installation is fine.

Note, my original message to this list with plain cygcheck.out
attached (as requested on the problems page) bounced for some reason
so I am trying a compressed version of cygcheck.out this time.

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
__________________________

[-- Attachment #2: compressed output from cygcheck for Cygwin on Wine --]
[-- Type: APPLICATION/x-gtar-compressed, Size: 3575 bytes --]

[-- Attachment #3: Type: text/plain, Size: 218 bytes --]

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Christopher Faylor]

On Fri, Jul 05, 2013 at 07:27:04PM -0700, Alan W. Irwin wrote:
>On 2013-07-05 16:48-0400 Christopher Faylor wrote:
>>On Fri, Jul 05, 2013 at 12:11:10PM -0700, Alan W.  Irwin wrote:
>>>On 2013-07-04 13:39-0700 Alan W.  Irwin wrote:
>>>>So if the consensus here after looking at the setup_install.out results
>>>>is that cygwin on Wine is more or less installed properly from these
>>>>results (except for the rename issue above which I can fix by doing
>>>>that rename manually as suggested by the error message), then the next
>>>>order of business [...]
>>>
>>>I have now looked more carefully at the setup_install.out results that
>>>are wrapped inside a tarball I attached to a previous post in this
>>>thread.  At the same time I have also looked at the resulting files in
>>>my cygwin install tree.  It turns out that because of errors the first
>>>install script created an empty /etc/fstab rather than the desired
>>>result and also was unable to create the /dev directory.
>>
>>The /dev directory is a pseudo-directory created on-the-fly by the
>>Cygwin DLL.
>
>I was really curious about the implications of that so had some further
>questions, but I am going to drop those questions because they are a
>distraction from the principal point I want to make.
>
>In sum, if any Cygwin developer here feels moved to help debug Wine so
>that the Cygwin install works properly on it, the best thing you could

I can only speak for this Cygwin developer but I have a long track
record of not wanting to involve myself in Cygwin and Wine interactions.
We stand on our head and do backflips in Cygwin code to get it to
emulate Linux behavior.  Having to do the same in another direction,
figuring out why something which works on multiple Windows platforms
does not work on a Windows emulator, is not something that I'm keen on
doing.

I'm just mentioning this in case you thought that since I have
sporadically been responding to this thread I might be interested in
joining in the discovery.  Unfortunately, I'm really not.

cgf

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Alan W. Irwin]

On 2013-07-05 16:48-0400 Christopher Faylor wrote:

> On Fri, Jul 05, 2013 at 12:11:10PM -0700, Alan W. Irwin wrote:
>> On 2013-07-04 13:39-0700 Alan W. Irwin wrote:
>> 
>>> So if the consensus here after looking at the setup_install.out
>>> results is that cygwin on Wine is more or less installed properly from
>>> these results (except for the rename issue above which I can fix by
>>> doing that rename manually as suggested by the error message), then
>>> the next order of business [...]
>> 
>> I have now looked more carefully at the setup_install.out results that
>> are wrapped inside a tarball I attached to a previous post in this
>> thread.  At the same time I have also looked at the resulting files in
>> my cygwin install tree.  It turns out that because of errors the first
>> install script created an empty /etc/fstab rather than the desired
>> result and also was unable to create the /dev directory.
> 
> The /dev directory is a pseudo-directory created on-the-fly by the
> Cygwin DLL.

I was really curious about the implications of that so had some further
questions, but I am going to drop those questions because they are a
distraction from the principal point I want to make.

In sum, if any Cygwin developer here feels moved to help debug Wine so
that the Cygwin install works properly on it, the best thing you could
do would be to follow the steps I detailed at at
<http://bugs.winehq.org/show_bug.cgi?id=24018>) with Wine-1.6-rc4
patched with Andrey Turkin's patch, and with Cygwin updated with the
recent snapshot cygwin1.dll that contains the fork fix.  Then run that
exact same setup.exe procedure (with the fork fix applied in the same
way) on Microsoft Windows.  Such direct comparisons between Wine and
Microsoft Windows results would be a great help in figuring out which
of the many error messages that show up in a fresh install of Cygwin
on Wine should actually be of concern to Wine developers.  (I cannot
do such comparisons myself because I don't have ready access to
Microsoft Windows, i.e., I have been running Linux since 1996 and my
only Windows experience is running MinGW/MSYS on Wine off and on as a
build and test platform for software for the last couple of years).

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
__________________________

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Christopher Faylor]

On Fri, Jul 05, 2013 at 12:11:10PM -0700, Alan W. Irwin wrote:
>On 2013-07-04 13:39-0700 Alan W. Irwin wrote:
>
>> So if the consensus here after looking at the setup_install.out
>> results is that cygwin on Wine is more or less installed properly from
>> these results (except for the rename issue above which I can fix by
>> doing that rename manually as suggested by the error message), then
>> the next order of business [...]
>
>I have now looked more carefully at the setup_install.out results that
>are wrapped inside a tarball I attached to a previous post in this
>thread.  At the same time I have also looked at the resulting files in
>my cygwin install tree.  It turns out that because of errors the first
>install script created an empty /etc/fstab rather than the desired
>result and also was unable to create the /dev directory.

The /dev directory is a pseudo-directory created on-the-fly by the
Cygwin DLL.

cgf

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Alan W. Irwin]

On 2013-07-04 13:39-0700 Alan W. Irwin wrote:

> So if the consensus here after looking at the setup_install.out
> results is that cygwin on Wine is more or less installed properly from
> these results (except for the rename issue above which I can fix by
> doing that rename manually as suggested by the error message), then
> the next order of business [...]

I have now looked more carefully at the setup_install.out results that
are wrapped inside a tarball I attached to a previous post in this
thread.  At the same time I have also looked at the resulting files in
my cygwin install tree.  It turns out that because of errors the first
install script created an empty /etc/fstab rather than the desired
result and also was unable to create the /dev directory. These issues
very likely had follow-on effects for several of the other postinstall
scripts that were run (for example, all the read-only filesystem error
messages when any attempt was made to write to /dev.) Anyhow, it now
seems likely there are at least a couple of Wine bugs that are still
preventing a clean postinstall phase for setup.exe on Wine, and the
rest of this story will continue for now at
http://bugs.winehq.org/show_bug.cgi?id=24018.  Meanwhile, I thank the
people here that (a) helped to fix the Cygwin fork bug, and (b) gave
me good directions for trying out that fix before it becomes part of
an official Cygwin release.

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
__________________________

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Failure with fork() (Cygwin on Wine install now sort of works)

[Alan W. Irwin]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 4867 bytes --]

Here is my take on the current status of Cygwin on Wine:

At http://bugs.winehq.org/show_bug.cgi?id=24018, Andrey Turkin took
advantage of the recent Cygwin fork fix to provide an additional Wine
fix (the small self-described "hackish" patch he gives there) that
allowed setup.exe to complete without run-time issues (albeit with
some non-zero return codes from the post-install scripts which I will
discuss in more detail below).

I would like to thank Achim Gratz for his suggestion in this thread of how to
correctly update cygwin1.dll in the middle of a setup.exe run.  I
have implemented his suggestion as the (MSYS or Linux) bash script
replace_cygwin1.dll.sh which is contained in the attached tarball
along with associated setup.exe results.

Here is how I used that script in practice

On Linux:

# Clean out previous Cygwin results completely
rm -rf cygwin cygwin_packages

# Clean startup of wine-1.6-rc4 built with patch mentioned above.
# Normally does nothing, but sometimes the -k option kills something
# that is left over from a previous wine run.
wineserver -k
# Start wine server
wineserver -p
# Start Wine in a bash environment
wineconsole --backend=curses MinGW-4.7.2/msys/1.0/bin/bash.exe

# From now on we are in the MSYS base on Wine environment.

# Set PATH to point to MSYS (so that the bash script below will run) then
# run
bash.exe-3.1$ ./setup.exe -n -N -d -R /z/home/wine/newstart/cygwin \
-l /z/home/wine/newstart/cygwin_packages \
-s ftp://mirror.cpsc.ucalgary.ca /cygwin.com -D >setup_download.out

bash.exe-3.1$ ./replace_cygwin1.dll.sh \
cygwin1-20130627.dll cygwin_packages

bash.exe-3.1$ ./setup.exe -n -N -d -R /z/home/wine/newstart/cygwin \
-l /z/home/wine/newstart/cygwin_packages \
-s ftp://mirror.cpsc.ucalgary.ca /cygwin.com -L >setup_install.out

The resulting setup_download.out and setup_install.out files are
contained in the attached tarball along with the
replace_cygwin1.dll.sh script.  I should also mention that
cygwin1-20130627.dll (which contains the latest fork fix) was
downloaded from the next to last Cygwin snapshot at
http://cygwin.com/snapshots/ and decompressed, and the version
of setup.exe that was run (downloaded a few days ago from
cygwin) is 2.774 according to its initial GUI.

I have used explicit setup.exe options above to be clear on exactly
what I did.  For all else available in the setup GUI, I just
took defaults such as not installing any additional packages, and
using the recommended system install rather than individual install.

The first "download" setup.exe invocation above had no obvious issues,
and similarly for the run of replace_cygwin1.dll.sh.  (For the latter
case I double checked that cygwin1-20130627.dll was identical to
cygwin/bin/cygwin1.dll, i.e, the script worked properly).

The second "install from local directory" setup.exe invocation above
completed without run-time issues.  (The first post-install script
hangs indefinitely without the Andrey Turkin patch to Wine mentioned
above.) However, there were 4 post-install scripts that generated
non-zero return codes.  Would someone take a look at setup_install.out
in the attached tarball to see how serious they are?  I have no idea
about that question myself.  They could be evidence of Wine issues (or
Cygwin issues that are also seen on Microsoft Windows) or they could
just be for the most part normal startup "chatter" when attempting to
install Cygwin for the very first time that will go away for
subsequent Cygwin use after this initial installation.  The only one
of the messages that seems obviously serious is

rebase: failed to rename "/etc/rebase.db.RIRMJ5" to "/etc/rebase.db.i386":
Invalid request code
Manually rename "/etc/rebase.db.RIRMJ5" to "/etc/rebase.db.i386",
otherwise the new rebase database will be unusable.

So if the consensus here after looking at the setup_install.out
results is that cygwin on Wine is more or less installed properly from
these results (except for the rename issue above which I can fix by
doing that rename manually as suggested by the error message), then
the next order of business is to attempt to actually use that shiny
new Cygwin install, but I will reserve my absolutely Cygwin newbie
questions about how to do that for a different thread.

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
__________________________

[-- Attachment #2: compressed tarball containing replace_cygwin1.dll.sh and setup.exe results --]
[-- Type: APPLICATION/x-gtar-compressed, Size: 44888 bytes --]

[-- Attachment #3: Type: text/plain, Size: 218 bytes --]

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple