Closed ceserthen closed 6 years ago
For the time being I modified line 169 of tumorcode/py/krebs/krebsjobs/submitVesselgeneration.py to if nums_points<4: to remove this error on the default case.
This is a bad idea since too small systems will crash the algorithm. In the release build this is checked only in python to speed things up. In order to obtain nicer c++ debug information build the code in the "DEBUG" version.
For this special issue increase the system size in order to obtain enough discretization points. This is also done by the exact command in the supplemental S1_example.pdf
s u b m i t V e s s e l g e n e r a t i o n −t 8 −p d e f a u l t −w 2000 − i 1 −e 1
I will also change this in the wiki now! Thanks for reporting.
Approach: fresh install
In a fresh install of ubuntu 16.04 without updates I was able to get submitVesselgeneration -t 8 -p defulat -w 2000 -i 1 -e
working using a Release
cmake build. On that system I am able to run submitVesselgeneration
, submitFakeTum
, and submitBulkTissue
without any problems but am still unable to run submitDetailedO2
as described in tutorial A.
Approach: original machine
Unfortunately on the original machine i am still unable to get the submitVesslgeneration to run. I used the submitVesselgeneration -p default -t 8 -w 2000
command with the num_points check restored to < 5. submitVesselgeneration was able to run, but still met returned the same segfault as before. I recompiled the code in debug mode (-DCMAKE_BUILD_TYPE=Debug
). Upon running the code I now receive the following traceback ImportError:
Traceback (most recent call last):
File "/home/donald/tc_install/bin/submitVesselgeneration", line 30, in <module>
import krebsutils
File "/home/donald/tc_install/py/krebsjobs/../krebsutils/__init__.py", line 43, in <module>
libkrebs = __import__('libkrebs_', globals(), locals())
ImportError: No module named libkrebs_
I'm clearing the cmake cache and respective directories and hoping that a new install will work for the second machine. Hopefuly I can find a simple fix to this problem.
Clearing the cache and simplifying the cmake script to the following appears to have solved the problem on the original machine. I also found a small issue with the submitDetailedO2
command referenced in Tutrorial A. submitDetailedO2 default vessels-default-typeI-15x19L130-sample00.h5 vessels
should instead be submitDetailedO2 default_o2 vessels-default-typeI-15x19L130-sample00.h5 vessels
to match paramterSetsO2.py
Again, I can only say thank your for taking the time to go thoroughly go trough the instructions and help to improve them. Yes, at some point I rename "default" to "default_o2" since I did not want so many defaults.
The ImportError: is hopefully not as severe as it appears since you probably have a library called "libkrebs_d.so" in your lib dir when compilling in debug mode. Sorry, but the debug options are not yet documented. The release "submit" scripts which are python wrappers do not yet support that. According to the python specs (see python -h
) you do have to invoke debug mode by adding python -d
flag. In your case this should be
python2 -d /home/donald/tc_install/py/krebsjobs/submitVesselgeneration.py -t 8 -p defulat -w 2000 -i 1
With the other scripts this works accordingly. The "-d" tells python to use the c++ library compilled with debug informations.
Since we work with the code across multiple linux distributions and also plan other architectures , the cmake has to be that elaborated.
While working through tutorial A during the vessel generation step. I first encountered an error with running
submitVesselgeneration -p default
giving a default number of points (4) giving the following errorFor the time being I modified line 169 of tumorcode/py/krebs/krebsjobs/submitVesselgeneration.py to
if nums_points<4:
to remove this error on the default case. I then ran thesubmitVesselgeneration -p default
and was given the following output:I then ran valgrind to look for any issues by running
valgrind mpirun submitVesselgeneration -p default
and received the following outputAny help resolving this issue is appreciated.