Hi. Today's silly question:
I'm trying to use SimulationTools to work with some Carpet HDF5 data. I've loaded SimulationTools & h5mma with no apparent issues, but this line fails:
ToDataRegion[ ImportHDF5[ "<full path to mydata.h5", {"Datasets", "mydataset"}]]
... with the error message:
LinkObject::linkd: Unable to communicate with closed link LinkObject['/Users/bjkelly1/Library/Mathematica/Applications/h5mma/MacOSX-x86-64/h5mma',397,6].
I can verify that the executable is where that error message says it is, and it's world-runnable. Any ideas why it's not working?
thanks
Beany
P.S. I've never used SimulationTools or h5mma before. I followed the simple installation instructions from http://simulationtools.org/Documentation/English/Tutorials/Introduction.html
BTW, I'm running Mathematica 10.4.0 on OS X 10.11.4 ("El Capitan")
B
On 12 April 2016 at 11:30, Bernard Kelly physicsbeany@gmail.com wrote:
Hi. Today's silly question:
I'm trying to use SimulationTools to work with some Carpet HDF5 data. I've loaded SimulationTools & h5mma with no apparent issues, but this line fails:
ToDataRegion[ ImportHDF5[ "<full path to mydata.h5", {"Datasets", "mydataset"}]]
... with the error message:
LinkObject::linkd: Unable to communicate with closed link LinkObject['/Users/bjkelly1/Library/Mathematica/Applications/h5mma/MacOSX-x86-64/h5mma',397,6].
I can verify that the executable is where that error message says it is, and it's world-runnable. Any ideas why it's not working?
thanks
Beany
P.S. I've never used SimulationTools or h5mma before. I followed the simple installation instructions from http://simulationtools.org/Documentation/English/Tutorials/Introduction.html
Hi Beany,
This suggests that there is a problem with the h5mma reader for HDF5 data. Can you try running the attached notebook in a fresh Mathematica kernel, changing the path to point to your own home directory? You should be able to reproduce the output that is in the notebook.
I have just run this on my Mac with Mathematica 10.4.0, OS X 10.11.4, SimulationTools 1.0.0 and h5mma 1.2.0 and it worked fine, so this should tell us if the problem is with h5mma, or elsewhere.
Regards, Barry
On Tue, Apr 12, 2016 at 4:41 PM, Bernard Kelly physicsbeany@gmail.com wrote:
BTW, I'm running Mathematica 10.4.0 on OS X 10.11.4 ("El Capitan")
B
On 12 April 2016 at 11:30, Bernard Kelly physicsbeany@gmail.com wrote:
Hi. Today's silly question:
I'm trying to use SimulationTools to work with some Carpet HDF5 data. I've loaded SimulationTools & h5mma with no apparent issues, but this line fails:
ToDataRegion[ ImportHDF5[ "<full path to mydata.h5", {"Datasets",
"mydataset"}]]
... with the error message:
LinkObject::linkd: Unable to communicate with closed link
LinkObject['/Users/bjkelly1/Library/Mathematica/Applications/h5mma/MacOSX-x86-64/h5mma',397,6].
I can verify that the executable is where that error message says it is, and it's world-runnable. Any ideas why it's not working?
thanks
Beany
P.S. I've never used SimulationTools or h5mma before. I followed the simple installation instructions from
http://simulationtools.org/Documentation/English/Tutorials/Introduction.html _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Hi Barry (and Ian). Thanks for your responses.
Barry, I tried the notebook you sent, and it fails in the same way (i.e. the bare ImportHDF5 command fails).
Ian, the HDF5 file *could* be corrupt, but I've read it into VisIt and used it with no issues.
As you suggested, I tried running h5mma on its own from a Terminal command line:
./h5mma
Create link:
... so it's expecting a "link", whatever it means by that. A target?
Beany
On 12 April 2016 at 12:17, Barry Wardell barry.wardell@gmail.com wrote:
Hi Beany,
This suggests that there is a problem with the h5mma reader for HDF5 data. Can you try running the attached notebook in a fresh Mathematica kernel, changing the path to point to your own home directory? You should be able to reproduce the output that is in the notebook.
I have just run this on my Mac with Mathematica 10.4.0, OS X 10.11.4, SimulationTools 1.0.0 and h5mma 1.2.0 and it worked fine, so this should tell us if the problem is with h5mma, or elsewhere.
Regards, Barry
On Tue, Apr 12, 2016 at 4:41 PM, Bernard Kelly physicsbeany@gmail.com wrote:
BTW, I'm running Mathematica 10.4.0 on OS X 10.11.4 ("El Capitan")
B
On 12 April 2016 at 11:30, Bernard Kelly physicsbeany@gmail.com wrote:
Hi. Today's silly question:
I'm trying to use SimulationTools to work with some Carpet HDF5 data. I've loaded SimulationTools & h5mma with no apparent issues, but this line fails:
ToDataRegion[ ImportHDF5[ "<full path to mydata.h5", {"Datasets", "mydataset"}]]
... with the error message:
LinkObject::linkd: Unable to communicate with closed link
LinkObject['/Users/bjkelly1/Library/Mathematica/Applications/h5mma/MacOSX-x86-64/h5mma',397,6].
I can verify that the executable is where that error message says it is, and it's world-runnable. Any ideas why it's not working?
thanks
Beany
P.S. I've never used SimulationTools or h5mma before. I followed the simple installation instructions from
http://simulationtools.org/Documentation/English/Tutorials/Introduction.html
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
On 12 Apr 2016, at 18:38, Bernard Kelly physicsbeany@gmail.com wrote:
Hi Barry (and Ian). Thanks for your responses.
Barry, I tried the notebook you sent, and it fails in the same way (i.e. the bare ImportHDF5 command fails).
Ian, the HDF5 file *could* be corrupt, but I've read it into VisIt and used it with no issues.
As you suggested, I tried running h5mma on its own from a Terminal command line:
./h5mma
Create link:
... so it's expecting a "link", whatever it means by that. A target?
That's what it's supposed to do. It's starting the protocol that it uses to communicate with Mathematica which usually launches it. So the executable can be loaded, but it looks like it is crashing.
There is test data provided with SimulationTools. Can you try that file?
ImportHDF5[FileNameJoin[{$SimulationToolsTestSimulation, "output-0000/bbh/phi.xy.h5"}]]
If that fails as well for you, and works for Barry, then that shows that the problem is not related to the file.
Barry, I assume you were testing with the development version? I looked at the changes since the release, and I don't see anything which might be a serious bug which has not been backported.
Thanks, Ian.
Tried the test data. Same error/failure.
I just checked out the git version of h5mma; the README seems mostly about the use of the utility in general, rather than installation instructions (I'm thinking of things like the prefix chosen at configuration stage). Should I build directly in the /Library/Mathematica/Applications directory (sounds wrong), or buld locally and move the result?
Beany
On 12 April 2016 at 12:49, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 12 Apr 2016, at 18:38, Bernard Kelly physicsbeany@gmail.com wrote:
Hi Barry (and Ian). Thanks for your responses.
Barry, I tried the notebook you sent, and it fails in the same way (i.e. the bare ImportHDF5 command fails).
Ian, the HDF5 file *could* be corrupt, but I've read it into VisIt and used it with no issues.
As you suggested, I tried running h5mma on its own from a Terminal command line:
./h5mma
Create link:
... so it's expecting a "link", whatever it means by that. A target?
That's what it's supposed to do. It's starting the protocol that it uses to communicate with Mathematica which usually launches it. So the executable can be loaded, but it looks like it is crashing.
There is test data provided with SimulationTools. Can you try that file?
ImportHDF5[FileNameJoin[{$SimulationToolsTestSimulation, "output-0000/bbh/phi.xy.h5"}]]
If that fails as well for you, and works for Barry, then that shows that the problem is not related to the file.
Barry, I assume you were testing with the development version? I looked at the changes since the release, and I don't see anything which might be a serious bug which has not been backported.
-- Ian Hinder http://members.aei.mpg.de/ianhin
On 12 Apr 2016, at 19:02, Bernard Kelly physicsbeany@gmail.com wrote:
Thanks, Ian.
Tried the test data. Same error/failure.
I just checked out the git version of h5mma; the README seems mostly about the use of the utility in general, rather than installation instructions (I'm thinking of things like the prefix chosen at configuration stage). Should I build directly in the /Library/Mathematica/Applications directory (sounds wrong), or buld locally and move the result?
There is a link in the README to the BitBucket wiki, which contains a link to the compilation instructions. This could be easier to find :)
https://bitbucket.org/simulationtools/h5mma/wiki/Compiling%20h5mma
On Mac OS, you shouldn't need to change any prefixes etc, if you are using homebrew or macports.
Hi Ian.
Thanks; I'm trying the compilation now, but am failing with a linking error I've encountered lately: "Undefined symbols for architecture x86_64". This seems to be an issue with MacPorts gcc compilers. I don't know if my MacPorts installation is borked, but I can't compile this right now.
... so any additional ideas on the precompiled h5mma binary are appreciated.
Beany
On 12 April 2016 at 13:08, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 12 Apr 2016, at 19:02, Bernard Kelly physicsbeany@gmail.com wrote:
Thanks, Ian.
Tried the test data. Same error/failure.
I just checked out the git version of h5mma; the README seems mostly about the use of the utility in general, rather than installation instructions (I'm thinking of things like the prefix chosen at configuration stage). Should I build directly in the /Library/Mathematica/Applications directory (sounds wrong), or buld locally and move the result?
There is a link in the README to the BitBucket wiki, which contains a link to the compilation instructions. This could be easier to find :)
https://bitbucket.org/simulationtools/h5mma/wiki/Compiling%20h5mma
On Mac OS, you shouldn't need to change any prefixes etc, if you are using homebrew or macports.
-- Ian Hinder http://members.aei.mpg.de/ianhin
On 12 Apr 2016, at 20:01, Bernard Kelly physicsbeany@gmail.com wrote:
Hi Ian.
Thanks; I'm trying the compilation now, but am failing with a linking error I've encountered lately: "Undefined symbols for architecture x86_64". This seems to be an issue with MacPorts gcc compilers. I don't know if my MacPorts installation is borked, but I can't compile this right now.
... so any additional ideas on the precompiled h5mma binary are appreciated.
Maybe you could try with homebrew?
The only thing I can think for the precompiled binary is to attach a debugger to it. h5mma should be loaded when you load SimulationTools; it will appear as a process in "ps". You could attach gdb:
/opt/local/libexec/gnubin/gdb --pid $(pgrep h5mma)
and then type "continue". Now, evaluate the ImportHDF5 expression from Mathematica. With any luck, you will get a backtrace in gdb when h5mma crashes. I wonder if the problem is related to the executable being compiled on an earlier version of OS X.
Nice idea.
The first thing I notice when loading h5mma is that gdb (now ggdb BTW, and I have to "sudo" for this purpose as well) generates an assload of Warnings of nonexistent files:
/Library/Caches/com.apple.xbs/Binaries/Metal/Metal-56.6~1/TempContent/Objects/Metal.build/Framework.build/Objects-normal/x86_64/MTLArgument.o [...]
In fact, the directory /Library/Caches/com.apple.xbs doesn't exist on my system. I wonder if this is related to my ultimate problem?
Moving on, the failure when using ImportHDF5 appears to exit "cleanly", leaving no backtrace:
------------------------------------------------ 0x00007fff92f9b07a in select$DARWIN_EXTSN () from /usr/lib/system/libsystem_kernel.dylib (gdb) continue Continuing. [Inferior 1 (process 40825) exited normally] (gdb) backtrace No stack. ------------------------------------------------
Presumably that first line might point to something, but I don't know what.
B
On 12 April 2016 at 14:11, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 12 Apr 2016, at 20:01, Bernard Kelly physicsbeany@gmail.com wrote:
Hi Ian.
Thanks; I'm trying the compilation now, but am failing with a linking error I've encountered lately: "Undefined symbols for architecture x86_64". This seems to be an issue with MacPorts gcc compilers. I don't know if my MacPorts installation is borked, but I can't compile this right now.
... so any additional ideas on the precompiled h5mma binary are appreciated.
Maybe you could try with homebrew?
The only thing I can think for the precompiled binary is to attach a debugger to it. h5mma should be loaded when you load SimulationTools; it will appear as a process in "ps". You could attach gdb:
/opt/local/libexec/gnubin/gdb --pid $(pgrep h5mma)
and then type "continue". Now, evaluate the ImportHDF5 expression from Mathematica. With any luck, you will get a backtrace in gdb when h5mma crashes. I wonder if the problem is related to the executable being compiled on an earlier version of OS X.
-- Ian Hinder http://members.aei.mpg.de/ianhin
Answering Ian's question: I tested with the release versions of SimulationTools (1.0.0) and h5mma (1.2.0) using the latest version of OS X (10.11.4) and Mathematica (10.4.0).
For debugging, you may have more success starting h5mma from the command line and then connecting to the executable from Mathematica. See https://bitbucket.org/simulationtools/h5mma/wiki/Debugging for instructions. As well as being able to use a debugger, this will enable you to see any standard output that may be printed by the HDF5 library.
On Tue, Apr 12, 2016 at 7:45 PM, Bernard Kelly physicsbeany@gmail.com wrote:
Nice idea.
The first thing I notice when loading h5mma is that gdb (now ggdb BTW, and I have to "sudo" for this purpose as well) generates an assload of Warnings of nonexistent files:
/Library/Caches/com.apple.xbs/Binaries/Metal/Metal-56.6~1/TempContent/Objects/Metal.build/Framework.build/Objects-normal/x86_64/MTLArgument.o [...]
In fact, the directory /Library/Caches/com.apple.xbs doesn't exist on my system. I wonder if this is related to my ultimate problem?
Moving on, the failure when using ImportHDF5 appears to exit "cleanly", leaving no backtrace:
0x00007fff92f9b07a in select$DARWIN_EXTSN () from /usr/lib/system/libsystem_kernel.dylib (gdb) continue Continuing. [Inferior 1 (process 40825) exited normally] (gdb) backtrace No stack.
Presumably that first line might point to something, but I don't know what.
B
On 12 April 2016 at 14:11, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 12 Apr 2016, at 20:01, Bernard Kelly physicsbeany@gmail.com wrote:
Hi Ian.
Thanks; I'm trying the compilation now, but am failing with a linking error I've encountered lately: "Undefined symbols for architecture x86_64". This seems to be an issue with MacPorts gcc compilers. I don't know if my MacPorts installation is borked, but I can't compile this right now.
... so any additional ideas on the precompiled h5mma binary are
appreciated.
Maybe you could try with homebrew?
The only thing I can think for the precompiled binary is to attach a debugger to it. h5mma should be loaded when you load SimulationTools; it will appear as a process in "ps". You could attach gdb:
/opt/local/libexec/gnubin/gdb --pid $(pgrep h5mma)
and then type "continue". Now, evaluate the ImportHDF5 expression from Mathematica. With any luck, you will get a backtrace in gdb when h5mma crashes. I wonder if the problem is related to the executable being compiled on an earlier version of OS X.
-- Ian Hinder http://members.aei.mpg.de/ianhin
On 12 Apr 2016, at 17:41, Bernard Kelly physicsbeany@gmail.com wrote:
BTW, I'm running Mathematica 10.4.0 on OS X 10.11.4 ("El Capitan")
Hi Beany,
This could indicate that h5mma has crashed, or possibly that it could never run it in the first place. Can you try running the h5mma executable from a terminal? It should give some output, rather than segfault.
Can you do an h5dump on the HDF5 file to check it is OK? If the file is corrupt, it's possible that the HDF5 library is segfaulting. You can also try reading it with the built-in Mathematica HDF5 reader. Just replace ImportHDF5 with Import. We found the built-in reader to be extremely flakey with complicated and large HDF5 files, but in simple cases it is probably fine.
B
On 12 April 2016 at 11:30, Bernard Kelly physicsbeany@gmail.com wrote:
Hi. Today's silly question:
I'm trying to use SimulationTools to work with some Carpet HDF5 data. I've loaded SimulationTools & h5mma with no apparent issues, but this line fails:
ToDataRegion[ ImportHDF5[ "<full path to mydata.h5", {"Datasets", "mydataset"}]]
... with the error message:
LinkObject::linkd: Unable to communicate with closed link LinkObject['/Users/bjkelly1/Library/Mathematica/Applications/h5mma/MacOSX-x86-64/h5mma',397,6].
I can verify that the executable is where that error message says it is, and it's world-runnable. Any ideas why it's not working?
thanks
Beany
P.S. I've never used SimulationTools or h5mma before. I followed the simple installation instructions from http://simulationtools.org/Documentation/English/Tutorials/Introduction.html
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
users@lists.einsteintoolkit.org
-
Barry Wardell -
Bernard Kelly -
Ian Hinder