Closed biagas closed 3 years ago
Hello,
could this be related to the issue I reported some time ago? https://github.com/visit-dav/live-customer-response/issues/157
In my case VisIt would not core dump, but the plot would not be produced and would get the error "[...] could not be generated by the compute engine on host "localhost"."
Cheer, AdV
Hello Vladimir,
When you say your software environment includes VTK-8.2, does that mean you have compiled VisIt from source against VTK-8.2? Did you use the build_visit script?
VisIt currently only supports and is known to work with VTK-8.1, so using 8.2 could be the source of the problem.
Are you able to work with a VisIt release tarball on your system? If not, can you rebuild VisIt with vtk-8.1?
Kathleen,
I do not use build_visit, in part because it did not work for me in the past, but mainly because it adds an extra layer above cmake, which should do the job on its own as designed. Basically, I proceed like this:
... a lot of patching applied to make the code to compile ... cmake -Wno-dev \ -DVISIT_PARALLEL:BOOL=ON \ -DVISIT_MPI_COMPILER:STRING="mpicc" \ -DVISIT_PYTHON_SCRIPTING:BOOL=OFF \ -DVISIT_PYTHON_FILTERS:BOOL=OFF \ -DVISIT_VTK_VERSION:STRING="8.2.0" \ -DVISIT_VTK_DIR:PATH=%{_prefix} \ -DVISIT_VTK_SKIP_INSTALL:BOOL=ON \ -DVISIT_QT_DIR:PATH=%{_prefix} \ -DVISIT_QT_SKIP_INSTALL:BOOL=ON \ -DVISIT_QWT_DIR:PATH=%{_prefix} \ -DVISIT_ZLIB_DIR:PATH=%{_prefix} \ -DVISIT_ZLIB_SKIP_INSTALL:BOOL=ON \ -DSILO_DIR:PATH=%{_prefix} \ -DHDF5_DIR:PATH=%{_prefix} \ -DNETCDF_DIR:PATH=%{_prefix} \ -DCMAKE_INSTALL_PREFIX:PATH=/opt/Visit make
(_prefix is /usr).
I have been using Visit 3.0.0 with VTK 8.2 for about a year with no problems. It is a major headache for me to try and build with software that is not native on my system. Ideally, I would prefer to figure out what causes the problem and fix it. I understand that it would also help the developers since the project will have to transition to a newer VTK sooner or later.
Thanks,
On Wed, Nov 18, 2020 at 10:10 AM Kathleen Biagas via visit-users < visit-users@elist.ornl.gov> wrote:
Hello Vladimir,
When you say your software environment includes VTK-8.2, does that mean you have compiled VisIt from source against VTK-8.2? Did you use the build_visit script?
VisIt currently only supports and is known to work with VTK-8.1, so using 8.2 could be the source of the problem.
Are you able to work with a VisIt release tarball on your system? If not, can you rebuild VisIt with vtk-8.1?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/visit-dav/live-customer-response/issues/179#issuecomment-729783134, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOLAYZFM7ZGMT4PI7LW2NFLSQPWXJANCNFSM4T2EYCCQ .
VisIt Users Wiki: http://visitusers.org/ Frequently Asked Questions for VisIt: https://wci.llnl.gov/simulation/computer-codes/visit/FAQ.html To Unsubscribe: send a blank email to visit-users-unsubscribe@elist.ornl.gov More Options: https://elist.ornl.gov/mailman/listinfo/visit-users
-- Vladimir Florinski
Yes, VisIt itself can be built quite successfully without the build_visit script. However, some thirdparty libaries (like VTK) need to be built with our script to ensure visit-centric patches and build settings are used.
Did you use the build_visit script to build third party libraries like VTK and QT? Or are you using pre-installed versions? Which version of QT and python are you using with VisIt?
The next VTK version VisIt will utilize will be in the 9.x series, the work to port VisIt to this version is a work-in-progress.
could this be related to the issue I reported some time ago? visit-dav/visit#12250
@angel-devicente, from our prespective, the two are related because they are built by means other than build_visit. IMHO, Spack is still experimental and building your own as Vladimir is doing makes it difficult for us to reproduce the behavior. We simply do not have the resources to support every possible configuration and so have to make choices.
@angel-devicente, from our prespective, the two are related because they are built by means other than build_visit. IMHO, Spack is still experimental and building your own as Vladimir is doing makes it difficult for us to reproduce the behavior. We simply do not have the resources to support every possible configuration and so have to make choices.
Of course, I understand. In my case I use Spack because the build_visit script was failing badly in my system (ArchLinux, the problems of living on the bleeding edge, I guess :-) ), but I thought maybe it was not a coincidence that in both cases Pseudocolors of regular variables worked no problem, but VisIt failed the moment we added expressions?
Vladimir,
Can you provide more information about the expressions you are creating that cause VisIt to crash?
@biagas... one thing I noticed about Vladimir's stack dump is that it appears to be happening when gathering the data on the engine to ship back to the viewer. I wonder...if the user set SR mode to Always
, would they still see this error?
Vladimir...sorry to be so cryptic in my comment here. What I mean is, go to Options->Renderin...
and follow these instructions to set Scalable Rendering
mode to Always
. See if that doesn't somehow work-around this bug.
Thank you for looking into my problem. Here is the extra information you requested.
First, my Visit installations do not make any changes to VTK. Some wrangling is required to get Visit to build, but the changes are miniscule and they don't affect VTK at all. For example, adding vtkRenderingGL2PSOpenGL2 library in one CMakeLists.txt or changing some member functions to use "unsigned char" arguments instead of "char". In my experience, Visit works reliably with stock VTK (currently with Fedora 33, but I have been using it from the time Fedora versions were in single digits).
Kathleen, I am not asking to support Visit on my system. It would be, of course, a waste of time for you and other developers. I am only asking for some analysis from the developers of what could be wrong and some guidance on to how to fix it. I will be looking forward to running Visit to work with stock VTK 9 once that is released for my system.
Second, any expression can cause the crash I reported. For example, I can define an expression as simply a copy of a database variable, and it would crash. I can even use built in expressions such as volume, Jacobian, etc., and it would crash also. However, everything works with database variables.
Third, enabling Scaled Rendering stopped the crashes. I hope this would provide some clue as to the nature of the problem.
Best,
On Wed, Nov 18, 2020 at 1:55 PM Mark C. Miller via visit-users < visit-users@elist.ornl.gov> wrote:
Vladimir...sorry to be so cryptic in my comment here. What I mean is, go to Options->Renderin... and follow these instructions https://visit-sphinx-github-user-manual.readthedocs.io/en/develop/gui_manual/Preferences/Rendering_Options_Window.html?highlight=scalable%20rendering#advanced-rendering-options to set Scalable Rendering mode to Always. See if that doesn't somehow work-around this bug.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/visit-dav/live-customer-response/issues/179#issuecomment-729917777, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOLAYZCHLYREZKG2EMXJZATSQQREPANCNFSM4T2EYCCQ .
VisIt Users Wiki: http://visitusers.org/ Frequently Asked Questions for VisIt: https://wci.llnl.gov/simulation/computer-codes/visit/FAQ.html To Unsubscribe: send a blank email to visit-users-unsubscribe@elist.ornl.gov More Options: https://elist.ornl.gov/mailman/listinfo/visit-users
-- Vladimir Florinski
Third, enabling Scaled Rendering stopped the crashes.
Ok, good to know. Thanks. That is because it winds up taking a different logic path during that data gathering step on the engine and avoiding whatever was causing the segv altogether. The downside is that you won't have GPU-accelerated rendering for the smaller datasets that might possibly benefit from it. I am sure when we upgrade to VTK-8.2 (or 9), we'll be able to reproduce and correct the problem then.
We're trying to combine
visit-users
email list with GitHub issues. When replying onvisit-users
, please reply only to emails with a Subject line that includes[visit-dav/live-customer-response]
.I am reporting the following behavior in Visit 3.1.3. The example discussed below pertains to 3D data stored in a SILO file, drawn using the Pseudocolor plot type. My software environment is Fedora 33 with VTK 8.2.
(1) Plot operators (Clip, Slice, etc.) work as expected with Pseudocolor plots with database variables (those defined in the SILO file), (2) User defined expressions can also be plotted, but only if no operators are applied, (3) Attempting to apply an operator to a Pseudocolor plot of a user defined expression dumps core
Here is a typical log:
Process 1055579 (engine_par) of user 1000 dumped core.
The error is evidently VTK related, but this is rather beyond my expertise. Would appreciate suggestions on how to fix this problem.
Thanks,
-- Vladimir Florinski