Matlab crashed when interpolating x, y coordinates to lat/long

JohannaBeckmann

HI so I installed ISSM on my Mac with the M1 chip. I am going through the tutorials and wanted to rund the GreenlandseaRISE experiment. when running I figured it crashes when it is interpolating x, y coordinates to lat/long the rest is fine.
because the warning told me to update my script I tried with the command (is that right?)
[md.mesh.lat,md.mesh.long]=CoordTransform(md.mesh.x,md.mesh.y,'EPSG:3413','EPSG:3413');
instead of the original xy2ll.
However, it still crashes and the terminal gives me:
bus error /Applications/MATLAB_R2022b.app/bin/matlab -r

I maybe not successfully installed proj? I don't know if I am running the right installation scripts here but in external packages/proj I run install-8-with_tests.sh and I get that the last 2 failed:
`skipped, 0 tests failed.
PROJ_SKIP_READ_USER_WRITABLE_DIRECTORY=YES PROJ_LIB=../../data/for_tests ../../src/gie deformation.gie
Reading file 'deformation.gie'
proj_create: Error 1029 (File not found or invalid): pipeline: Pipeline: Bad step definition: proj=deformation (File not found or invalid)
proj=pipeline step proj=cart ellps=GRS80 step proj=deformation grids=t...
Invalid operation definition in line no. 36:
File not found or invalid (errno=invalid_op_file_not_found_or_invalid/1029)
proj_create: Error 1026 (Missing argument): deformation: Either +grids or (+xy_grids and +z_grids) should be specified.
proj_create: Error 1026 (Missing argument): deformation: Either +grids or (+xy_grids and +z_grids) should be specified.
proj_create: Error 1029 (File not found or invalid): deformation: could not find requested xy_grid(s).
proj_create: Error 1029 (File not found or invalid): deformation: could not find requested z_grid(s).

proj_create: Error 1029 (File not found or invalid): deformation: could not find requested z_grid(s).

total: 39 tests succeeded, 0 tests skipped, 2 tests FAILED!

make[5]: *** [deformation] Error 2
make[4]: *** [check-am] Error 2
make[3]: *** [check-recursive] Error 1
make[2]: *** [check-local] Error 2
make[1]: *** [check-am] Error 2
make: *** [check-recursive] Error 1`

Ignored and moved forward and I put in my configure script

the line:
--with-proj-dir="$ISSM_DIR/externalpackages/proj/install" \

do the
'autoreconf -ivf
make
make install'
which seems to work but I guess not really?
thanks

justinquinn

JohannaBeckmann
Hi Johanna,
Sorry to hear you're running into problems. The following is the list of packages and installation scripts we use for our own Silicon-based Mac build node,

autotools	install-mac.sh
cmake		install.sh
petsc		install-3.17-mac-silicon.sh
gsl			install.sh
boost		install-1.7-mac.sh
dakota		install-6.2-mac.sh
curl		install-7-mac.sh
hdf5		install-1-parallel.sh
netcdf		install-4.7-parallel.sh
proj		install-6.sh
gdal		install-3-python.sh
gshhg		install.sh
gmt			install-6-mac.sh
gmsh		install-4-mac.sh
triangle	install-mac.sh
chaco		install-mac.sh
m1qn3		install.sh
semic		install.sh

You may not need all of the above (in particular, Boost and Dakota), but I figured I would include everything for the sake of completeness. Your configuration would then include,

--with-metis-dir=${ISSM_DIR}/externalpackages/petsc/install \
--with-parmetis-dir=${ISSM_DIR}/externalpackages/petsc/install \
--with-scalapack-dir=${ISSM_DIR}/externalpackages/petsc/install \
--with-mumps-dir=${ISSM_DIR}/externalpackages/petsc/install \
--with-hdf5-dir=${ISSM_DIR}/externalpackages/hdf5/install \
--with-petsc-dir=${ISSM_DIR}/externalpackages/petsc/install \
--with-gsl-dir=${ISSM_DIR}/externalpackages/gsl/install \
--with-boost-dir=${ISSM_DIR}/externalpackages/boost/install \
--with-dakota-dir=${ISSM_DIR}/externalpackages/dakota/install \
--with-proj-dir=${ISSM_DIR}/externalpackages/proj/install \
--with-triangle-dir=${ISSM_DIR}/externalpackages/triangle/install \
--with-chaco-dir=${ISSM_DIR}/externalpackages/chaco/install \
--with-m1qn3-dir=${ISSM_DIR}/externalpackages/m1qn3/install \
--with-semic-dir=${ISSM_DIR}/externalpackages/semic/install \

again, removing any configuration options that you are not using.

I would start by trying to reinstall PROJ with the install-6.sh script, then reconfiguring and reinstalling ISSM. Please let us know if that helps.

JohannaBeckmann

Hi Justin,
Thanks for the fast reply. However, I had no success with version 6. It still crashes.
same error message:
zsh: bus error /Applications/MATLAB_R2022b.app/bin/matlab -r
before the transformation in step 1 I have this output:
Anisotropic mesh adaptation WARNING: mesh present but no geometry found. Reconstructing... new number of triangles = 6345

justinquinn

JohannaBeckmann Can you try switching to bash as your shell? I wonder if source $ISSM_DIR/etc/environment.sh is not working as expected because you are under zsh?

JohannaBeckmann

justinquinn
so this is all in my .zshrc , I put an echo command into the environment.sh and when opening the terminal I see the the echo output. so that should be saying it is actually working right? or should I still try to switch to bash? I tried copying to .zshrc into .bashrc however my aliases don"t work. how do I tell the bash terminal to checkout .bashrc?

justinquinn

JohannaBeckmann Put it all instead in ~/.bash_profile and, yes, still try with bash as your shell.

JohannaBeckmann

ok , switched, crashed also reinstalled , crashed again, but now with this error message:
Bus error: 10
Is that any good?

justinquinn

JohannaBeckmann Sounds like it's just about the same error, which I haven't seen before. I am going to tag @mathieumorlighem in on this to take a look when he gets a chance. My guess is that it's either (1) something off in your call to CoordTransform, or, (2) we actually have a segmentation fault somewhere in our core code. Thanks for your patience on this.
If I might ask, when did you download the binaries from our website? We just generated new ones about a week ago, so it may be worth trying to download a new copy to see if that changes anything.

JohannaBeckmann

justinquinn
OK thanks Justin, let's hope for the better.
I just dowloaded the binaries and replaced them into my lib2 folder, recompiled but sadly no progress 🙁
Also updated ISSM, no progress either (also no big changes) It is the official release and not the developer branch.

mathieumorlighem

Hi Johanna!

Good to see you here! gdal/proj is a bit of a nightmare, and this looks like a conflict of library with MATLAB's own libs which is not easy to debug... we restored xy2ll and ll2xy to their original (native) version. We can try to find a workaround for you, but in the meantime, can I ask why you are trying to use CoordTransform and not xy2ll ?
Thanks
Mathieu

justinquinn

mathieumorlighem That's a good point about a conflict with MATLAB's libraries. @JohannaBeckmann, after running source $ISSM_DIR/etc/environment.sh, what do you get from running ldd matlab (if you do not have MATLAB on your PATH, use the full path to matlab in this command).

JohannaBeckmann

mathieumorlighem
HI Mathieu, It comes as an output in the Matlab while I am running step 1.
so with the original script an d xy2ll I get:
Calling CoordTransform instead, make sure to change your MATLAB script
Because it crashed, I thought , well I better do that then 🙂
but both commands crash (or coordTransform is called every time?)
Well, I think I could do do the projection transform outside of Matlab, but I thought , just one line here looks so easy and promising and it would be great if that works.

JohannaBeckmann

justinquinn

I tried to translate that command into OS but
otool -L /Applications/MATLAB_R2022b.app/bin/matlab

gives me

/Applications/MATLAB_R2022b.app/bin/matlab: is not an object file

mathieumorlighem

JohannaBeckmann Hi Johanna

Alright, it looks like you have the old version of ISSM, we may not have pushed the latest changes to the public branche. One thing you can do to fix this is to simply delete CoordTransform_matlab.mexmaci64 (Double check the extension though, in MATLAB if you type which CoordTransform_matlab, it will show you the file you need to delete). Sorry about this hiccup, everything else should work fine 🙂
Mathieu

justinquinn

JohannaBeckmann Yes, my bad: ldd would be on a Linux system. I thought that otool -L would work on an executable, but apparently does not. Anyhow, it seems that Mathieu has you on the right path by updating those modules.

JohannaBeckmann

Thanks it works fine now!