#2039: Startup Error
----------------------+-----------------------------------------------------
Reporter: akdeakin | Owner: grass-dev@…
Type: defect | Status: new
Priority: normal | Milestone: 6.4.3
Component: Startup | Version: 6.4.3 RCs
Keywords: | Platform: MSWindows 7
Cpu: x86-32 |
----------------------+-----------------------------------------------------
GRASS GIS 6.4.3RC2 was downloaded as part of Quantum GIS. When starting
GRASS, command line window appears with multiple errors:
ERROR: G_getenv(): Variable LOCATION_NAME not set
I've attached a jpeg of the command window. The window appears briefly
before closing and there's a very brief display of opening graphic.
I am running this on my computer at work where I don't have administrative
privileges. All installs, uninstalls, etc. have to be done via a technical
support request. I wasn't sure whether not having administrative
privileges may have anything to do with the errors.
The sample data (e.g., Spearfish60) are stored in
c:\Users\deakin\grassdata.
A series of wxGUI related errors have been fixed since your
installed RC2 version. And, yes, administrator rights are needed
to install the software (not to run it, for this normal user rights
are the proper way to use the software).
I have upgraded to GRASS GIS 6.4.3RC4:
WinGRASS-6.4.4svn-r57305-650-Setup.exe
This time, I am trying to launch GRASS on my home computer with full
administrative privileges.
I have also installed to the following directory: C:\grass64 and
installed the latest Microsoft .NET framework. I also disabled my avast
antivirus software. I have reinstalled the MSVCR71.dll to both the
System32 and SysWOW64 folders
When I try to launch GRASS GIS 6.4.4svn, a command line window flashes
very quickly and then nothing.
Do you have any further things I might try? Thanks for your patience and
help.
it sounds like you have most of the runtime redistributables in place,
msvcr100.dll is a new one though. When you installed did you tick the box
to install "Important Microsoft Redistributable Libraries" on the same
page in the installer as the option to download the sample datasets? Even
though you probably have most already it might be a good idea to try that
anyway.
To see what the error message is you have to stop the window from closing
so fast once it exits. Perhaps the easiest way around that is to open a
command window (start button then type "cmd" into the search and click in
the "C:\>" icon) and from the DOS prompt change into the grass startup
directory then run the startup command by hand.
To make this easy right click on the grass desktop icon and copy and paste
out the "start from" directory to the terminal with "cd
c:\path\goes\here", I think that points to the c:\grass64\msys dir (for
you) but I'm not sure it actually matters.
Then cut and paste the actual startup command into the command terminal,
and wait to see the error.
Is it still "LOCATION_NAME not set"? (which should be recoverable) Do any
of the other grass session menu items work from the grass menu in start
menu -> all programs? For example, if you can start from the text based
command session, maybe that fixes the broken/missing "grassrc6" last-
settings file which remembers the LOCATION_NAME from the previous session.
?
Thank so much for your detailed suggestions. They are much appreciated!
I did re-download and "install" msvcr100.dll. I uninstalled Grass and re-
installed it, making sure to tick the box next to Important Microsoft
Redistributable Libraries" on the same page in the installer as the option
to download the sample datasets.
While I am able to successfully run the Old TclTk Gui and Command Line
options, I'm still unable to run the new GUI option.
From the DOS command line, when I run grass64svn I get 2 instances of
ImportError: cannot import name MAXREPEAT. Both are preceded by calls to
the Python27 library. (E.g., File "C:\grass64\Python27\lib\site.py", line
548 in <module> main <>)
So, I upgraded from Python 2.7.4 to 2.7.5, but I'm getting the same
response.
Given that I can access Grass through the Old TclTk Gui, this isn't a
priority...I'm just curious, now.
Thanks, again. I really appreciate your help and your rapid response.
Replying to [comment:4 akdeakin]:
> Hamish,
>
> Thank so much for your detailed suggestions. They are much appreciated!
>
> I did re-download and "install" msvcr100.dll. I uninstalled Grass and
re-installed it, making sure to tick the box next to Important Microsoft
Redistributable Libraries" on the same page in the installer as the option
to download the sample datasets.
>
> While I am able to successfully run the Old TclTk Gui and Command Line
options, I'm still unable to run the new GUI option.
>
> From the DOS command line, when I run grass64svn I get 2 instances of
ImportError: cannot import name MAXREPEAT. Both are preceded by calls to
the Python27 library. (E.g., File "C:\grass64\Python27\lib\site.py", line
548 in <module> main <>)
>
> So, I upgraded from Python 2.7.4 to 2.7.5, but I'm getting the same
response.
This had happened before and was diagnosed at the time as a Python issue
[1], but with version Python < 2.4. Are you sure that the Python used by
your GRASS installation is the system wide one ? It might a Python version
embedded in QGIS ?
To make sure, if that is possible for you, you might want to try to
uninstall all Python installations on your machine + QGIS + GRASS and then
reinstall only GRASS from the standalone GRASS installer.
For my colleagues this problem appears for GRASS7svn and I believe that
also for GRASS64svn. It does not seem to be a problem with GRASS6.4.3RC*
installed through Helmut's installer. It appears that ADK has been using
the svn snapshots - do you still have this problem if you try the latest
GRASS6.4.3RC4 from here? http://grass.osgeo.org/grass64/binary/mswindows/native/
For the record, this is what my colleagues found (which indicates python
version conflict with ArcGIS, already discussed on the list): GRASS 7
needs Python 2.7. I have several Python 2.7 installed on my machine, so
we started looking at that.
Now Richard has found that GRASS 7 works perfectly on Windows 7 and 8
until ArcGIS 10.1 is installed. Then, when ArcGIS 10.1 is installed,
GRASS 7 exits immediately upon starting. Actually, if you pause this
jing video at 6 seconds, it shows 'importError: cannot import name
MAXREPEAT'
We have tried, unsuccessfully, the fixes we could find online for the
MAXREPEAT error. We were wondering if you could communicate this with the
developers to see if they have some ideas about how to tackle this
problem.
Replying to [comment:6 helena]:
[...]
> Now Richard has found that GRASS 7 works perfectly on Windows 7 and 8
until ArcGIS 10.1 is installed. Then, when ArcGIS 10.1 is installed,
GRASS 7 exits immediately upon starting. Actually, if you pause this
jing video at 6 seconds, it shows 'importError: cannot import name
MAXREPEAT'
confirmed.
ArcGIS 10.0 used C:\'''Python26'''\ArcGIS10.0
ArcGIS 10.1 uses C:\'''Python27'''\ArcGIS10.1
Replying to [comment:6 helena]:
...
> For the record, this is what my colleagues found (which indicates python
version conflict with ArcGIS, already discussed on the list): GRASS 7
needs Python 2.7. I have several Python 2.7 installed on my machine, so
we started looking at that.
> Now Richard has found that GRASS 7 works perfectly on Windows 7 and 8
until ArcGIS 10.1 is installed. Then, when ArcGIS 10.1 is installed,
GRASS 7 exits immediately upon starting. Actually, if you pause this
jing video at 6 seconds, it shows 'importError: cannot import name
MAXREPEAT'
Q: Error message "!ImportError: cannot import name MAXREPEAT"
A: A system-wide Python installation (2.7.2, independent of GRASS) may
cause the trouble. Replacing this system-wide Python installation with
version 2.7.4, instead of 2.7.2, solved the problem.
Replying to [comment:8 neteler]:
> Replying to [comment:6 helena]:
> ...
> > For the record, this is what my colleagues found (which indicates
python version conflict with ArcGIS, already discussed on the list): GRASS
7 needs Python 2.7. I have several Python 2.7 installed on my machine,
so we started looking at that.
> > Now Richard has found that GRASS 7 works perfectly on Windows 7 and 8
until ArcGIS 10.1 is installed. Then, when ArcGIS 10.1 is installed,
GRASS 7 exits immediately upon starting. Actually, if you pause this
jing video at 6 seconds, it shows 'importError: cannot import name
MAXREPEAT'
>
> Perhaps this helps (http://grasswiki.osgeo.org/wiki/WinGRASS_errors):
>
> Q: Error message "!ImportError: cannot import name MAXREPEAT"
>
> A: A system-wide Python installation (2.7.2, independent of GRASS) may
cause the trouble. Replacing this system-wide Python installation with
version 2.7.4, instead of 2.7.2, solved the problem.
Could someone check what exact version ArcGIS installs. I suspect < 2.7.4.
Unfortunately, the Python launcher discussed in the thread Helmut points
to in comment:7 seems to only take into account minor version numbers, but
not revision numbers, so there does not seem to be a way to distinguish
between two versions of 2.7 on the machine.
Is it possible to uninstall the python installed by ESRI ?
One option would probably be to delete the registry key entries of the
ESRI Python version, install a more recent Python (e.g. 2.7.5) and install
the Python launcher.
{{{
=> it's the system wide python 2.7.2 installed by ArcGIS10.1.
}}}
>
> Unfortunately, the Python launcher discussed in the thread Helmut points
to in comment:7 seems to only take into account minor version numbers, but
not revision numbers, so there does not seem to be a way to distinguish
between two versions of 2.7 on the machine.
>
> Is it possible to uninstall the python installed by ESRI ?
only if you deinstall ArcGIS ...
>
> One option would probably be to delete the registry key entries of the
ESRI Python version, install a more recent Python (e.g. 2.7.5) and install
the Python launcher.
but then ArcGIS is unusable ... ok, it's not open source ... ;o)
Replying to [comment:10 hellik]:
> >
> > Unfortunately, the Python launcher discussed in the thread Helmut
points to in comment:7 seems to only take into account minor version
numbers, but not revision numbers, so there does not seem to be a way to
distinguish between two versions of 2.7 on the machine.
>
> see also http://trac.osgeo.org/grass/ticket/2015#comment:6
just forgotten, according to the launcher manual it should be possible.
Replying to [comment:11 hellik]:
> Replying to [comment:10 hellik]:
> > >
> > > Unfortunately, the Python launcher discussed in the thread Helmut
points to in comment:7 seems to only take into account minor version
numbers, but not revision numbers, so there does not seem to be a way to
distinguish between two versions of 2.7 on the machine.
> >
> > see also http://trac.osgeo.org/grass/ticket/2015#comment:6
>
> just forgotten, according to the launcher manual it should be possible.
Replying to [comment:10 hellik]:
> Replying to [comment:9 mlennert]:
> >
> > Could someone check what exact version ArcGIS installs. I suspect <
2.7.4.
>
> have a look at http://trac.osgeo.org/grass/ticket/2015#comment:6
>
> {{{
> => it's the system wide python 2.7.2 installed by ArcGIS10.1.
> }}}
>
> >
> > Unfortunately, the Python launcher discussed in the thread Helmut
points to in comment:7 seems to only take into account minor version
numbers, but not revision numbers, so there does not seem to be a way to
distinguish between two versions of 2.7 on the machine.
>
> see also http://trac.osgeo.org/grass/ticket/2015#comment:6
>
> >
> > Is it possible to uninstall the python installed by ESRI ?
>
> only if you deinstall ArcGIS ...
>
> >
> > One option would probably be to delete the registry key entries of the
ESRI Python version, install a more recent Python (e.g. 2.7.5) and install
the Python launcher.
>
> but then ArcGIS is unusable
Is this because of a specific pythonpath they register ?
AFAIK we use `"$GRASS_PYTHON" "$SCRIPT.py"` everywhere now, and as long as
we didn't miss any, so registry associations with other system Python
installations shouldn't matter since python.exe is called with its full
path. the trouble might be instead that the order of the PYTHONPATH is
picking up the alternate install before the grass install. You might try
editing the lines of the python code just before the traceback's line
number and add a "print sys.path" to show the path order. Right now we are
adding the grass paths to the end of that, may we need to be inserting
them at the start of the list instead.
Replying to [comment:15 hamish]:
> AFAIK we use `"$GRASS_PYTHON" "$SCRIPT.py"` everywhere now, and as long
as we didn't miss any, so registry associations with other system Python
installations shouldn't matter since python.exe is called with its full
path.
This works for the GUI, but $GRASS_PYTHON is not usable when calling a
module from the command line. This is why calling python scripts from the
command line does not work in Windows. The same is true for scripts
calling scripts.
This is why we started looking into the Python launcher as a solution, and
the Python launcher does seem to be using the registry, not only for file
association, but also for the definition of PYTHONPATH.
> the trouble might be instead that the order of the PYTHONPATH is picking
up the alternate install before the grass install. You might try editing
the lines of the python code just before the traceback's line number and
add a "print sys.path" to show the path order. Right now we are adding the
grass paths to the end of that, may we need to be inserting them at the
start of the list instead.
In my installation, when I try echo %PYTHONPATH% on the command line, I
get the GRASS paths at the beginning.
The discussion continued in another ticket (#1941). It seems that Martin
solved the MAXREPEAT problem (do we need to test more?). It's not clear to
me if the original problem in description is also solved?
Replying to [comment:17 annakrat]:
> The discussion continued in another ticket (#1941). It seems that Martin
solved the MAXREPEAT problem (do we need to test more?). It's not clear to
me if the original problem in description is also solved?
So there is another MAXREPEAT problem. When you run g.extension wrapper or
in python console:
Replying to [comment:18 annakrat]:
> Replying to [comment:17 annakrat]:
> > The discussion continued in another ticket (#1941). It seems that
Martin solved the MAXREPEAT problem (do we need to test more?). It's not
clear to me if the original problem in description is also solved?
>
> So there is another MAXREPEAT problem. When you run g.extension wrapper
or in python console:
>
> {{{
> grass.read_command('g.extension.py', flags='l')
> }}}
>
> you get nothing because of the MAXREPEAT error. But at least you can
still install extension through the the module dialog and command line.
