Hello,
We just had one of those "PyEval_RestoreThread: NULL tstate" errors
cropping up on Linux, just before this release. It was when you tried to
go to a position in a window which had not yet been opened. It turned out
not to be a timing issue. That is, the window has not yet been fully
initialised (as far as Tk was concerned) before Analysis was trying to
change its position. We worked around that one by delaying the latter
change with an "after_idle". I wouldn't be surprised if something similar
is happening on the Mac.
Can you send us the XML file which is causing this problem (not the data
files)? We have a Mac here with (just about) 1.0.4 installed and it is
not causing that problem with multiple spectra and windows. (It seems to
be causing other problems, e.g. the screens are not always refreshed the
way they should.)
Thanks, Wayne
On Wed, 29 Jun 2005, Bruce D. Ray wrote:
> The system is a G4 running OS X 10.3.9 with the fink installations
> of python 2.3 with numeric 23.1, tcl/tk 8.4, and glut 3.7 (added
> as a quick fix to circumvent compilation problems with analysis 1.02,
> and not actually used).
>
> Compilation did not include use of OpenGL, but was for tcl/tk graphics
> only. Previously compiled analysis 1.02 with this configuration worked
> without any difficulties at all. A series of messages in May about
> analysis 1.0.3, document a python fatal error on attempts to open a
> previously saved tutorial at the command line.
>
> The build proceeded as usual with the standard warnings about the
> deprecation of and possible future replacement of the long double type.
>
> As with analysis 1.0.3, I get an error message on some attempts to open
> a previously saved tutorial project on the command line as follows:
>
> analysis test1.xml
> CCPNMR Analysis Version 1.0. Release 4 (Copyright 2003-2004 CCPN)
> Distribution created Wed Jun 29 08:12:22 2005
> >>> Fatal Python error: PyEval_RestoreThread: NULL tstate
> /ccpnmr/bin/analysis: line 1: 1176 Abort trap $CCPNMR_TOP_DIR/bin/python -i -O $CCPNMR_TOP_DIR/ccpnmr1.0/python/ccpnmr/analysis/AnalysisGui.py $1 $2 $3
>
>
> I've not seen 1176 given as the Abort trap number before, but I don't
> know the significance of that number. At the time of the error, analysis
> has just started opening the second spectrum graphics window which contains
> two strips with assignments and which serves both of the 3-D spectra of
> the tutorial. I suspect there is some multi-threading going on so that
> the display's indications of timing do not reflect the actual timing.
> If the X11 session is freshly started and "analysis test1.xml" is the
> first graphics requiring command given, the tutorial project opens
> correctly. The abort only happens after analysis has been used once,
> has been quit, and is started again in the same X11 session. Could
> this be an indication of memory release or memory leak problems?
>
> Despite this error message with the saved tutorial project, analysis 1.0.4
> does open from the command line a working project with a spectral single
> graphics window that serves 2-D spectra without difficulties.
>
>
> Sincerely,
>
>
> --
> Bruce D. Ray, Ph.D.
> Associate Scientist, and Operations Director
> NMR Center
> IUPUI
> Physics Dept.
> 402 N. Blackford St.
> Indianapolis, IN 46202-3273
>
|