search

LLNL / paraDIS_lib

Visualization library to work with the paraDIS dislocation dynamics code
Other
11 stars 4 forks source link

Segmentation fault #1

Open shiro-ihara opened 6 years ago

shiro-ihara commented 6 years ago

Hello!

When I typed "./bin/analyzeParaDIS --full rs0005.data" the output was "_Segmentation fault_". What should I do?

For your information, rs0005.data is the restart file of paraDIS, and the gcc-version is as follows. gcc (GCC) 7.2.1 20170915 (Red Hat 7.2.1-2) Boost and Tclap are already installed.

wealthychef1 commented 6 years ago

Hi, Good to know that analyzeParaDIS is still in circulation and use. :-) The --full flag will create a lot of debug output, so the first step is to read the debug files to see where it seg faulted.
Of course the code should never segfault, but only you understand your data, so it's impossible for me to say why this has happened.
Warning: this kind of error can take days to find, and is almost always because of some unforeseen oddity in the data set that reflects a physical impossibility in the simulation. I'm surprised it happened at timestep 5.

I apologize but I am retired now and do not use this code any more. I do think that you can find the problem using the debug files it leaves behind; they are very very detailed.

Hope this helps!

-- Rich

On Feb 5, 2018, at 2:49 PM, shiro-ihara notifications@github.com wrote:

Hello!

When I typed "./bin/analyzeParaDIS --full rs0005.data" the output was "Segmentation fault". What should I do?

For your information, rs0005.data is the restart file of paraDIS, and the gcc-version is as follows. gcc (GCC) 7.2.1 20170915 (Red Hat 7.2.1-2) Boost and Tclap are already installed.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/LLNL/paraDIS_lib/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AEmhaJ9FoOTpK4aU_flbi4bBJ-WR55Srks5tRrJugaJpZM4R5CNd.

Rich Cook

It's not the time it takes to take the take that takes the time; it's the time it takes between the takes that takes the time. - Roy Scheider

shiro-ihara commented 6 years ago

Thank you for your kindness.

I want to read the debug files, but cannot find it. I think it can be found by searching the directly, but I couldn't. Where is it?

wealthychef1 commented 6 years ago

Hi, It will be in the same directory as the data I think. I thought it would tell you.
I don't have analyzeParaDIS built on my mac at the moment; should do that but it failed and I'm going out in a minute...

Maybe try the --help flag on analyzeParaDIS. I think there is a --verbose flag too, I thought that was automatic with --full but maybe not.
Or run analyzeParaDIS with no arguments, that should give you a help menu too.

Maybe something like this will work: analyzeParaDIS --verbose 3 --full rs005.dat

On Feb 5, 2018, at 4:22 PM, shiro-ihara notifications@github.com wrote:

Thank you for your kindness.

I want to read the debug files, but cannot find it. I think it can be found by searching the directly, but I couldn't. Where is it?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/LLNL/paraDIS_lib/issues/1#issuecomment-363025888, or mute the thread https://github.com/notifications/unsubscribe-auth/AEmhaOX2Pc108mlx_hROMJQvZV_Pjyokks5tRshAgaJpZM4R5CNd.

Rich Cook

It's not the time it takes to take the take that takes the time; it's the time it takes between the takes that takes the time. - Roy Scheider

shiro-ihara commented 6 years ago

Hello, I tried to run analyzeParaDIS with no arguments, but the result was Segmentation fault. The debug file was also not found.

Maybe, my environmental setting is something wrong. I will check it. Thank you

wealthychef1 commented 6 years ago

That is interesting. Do you have a debugger?
It now sounds like it's not a data problem but rather a build problem linking libraries. Sometimes during the build, some headers get included from one version of a library and then the actual library that gets linked is different, and this causes a segfault as the data structures are different in the library than in the headers.
A common source of this is the so-called boost libraries. -- Rich

On Feb 6, 2018, at 9:04 AM, shiro-ihara notifications@github.com wrote:

Hello, I tried to run analyzeParaDIS with no arguments, but the result was Segmentation fault. The debug file was also not found.

Maybe, my environmental setting is something wrong. I will check it. Thank you

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/LLNL/paraDIS_lib/issues/1#issuecomment-363286014, or mute the thread https://github.com/notifications/unsubscribe-auth/AEmhaAoTo5Yv-az2gnWLoKTBanrTi3JNks5tR7MPgaJpZM4R5CNd.

Rich Cook

It's not the time it takes to take the take that takes the time; it's the time it takes between the takes that takes the time. - Roy Scheider

shiro-ihara commented 6 years ago

I didn't have a debugger, so I installed it. But my computer still send the message "Segmentation fault". The debug files seem to be not created. I have built boost libraries and set BOOSTDIR, which in the makefile, to reach it. Tclap is too . Is there something I should edit more?

wealthychef1 commented 6 years ago

For the debugger, merely installing it is not enough. You have to actually run the program under the debugger. When the program crashes, the debugger will show you where it crashed. This is not always easy to understand.

As a guess, I suspect you have two copies of boost or tclap or...? installed. The one that the makefile builds against and the one that is in your LD_LIBRARY_PATH

If it's boost, then it might look like this:

build links against /my/boostlib/lib_boost. http://lib_boost.la/so at runtime, analyzeParaDIS loads /usr/local/lib/lib_boost.so

You can discover which analyzeParaDIS is loading by running ldd like this: ldd analyzeParaDIS

You will get output that shows the exact libraries being loaded at runtime. When you compare this to the debugger, it will tell you perhaps the problem.

I don't have analyzeParaDIS built but here is an example of "ldd b2m"

rcook@linode ~ $ ldd /usr/local/bin/b2m linux-vdso.so.1 (0x00007ffc92ba6000) libncurses.so.5 => /lib/x86_64-linux-gnu/libncurses.so.5 (0x00007f9cf0148000) libtinfo.so.5 => /lib/x86_64-linux-gnu/libtinfo.so.5 (0x00007f9ceff1e000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f9cefc1d000) libutil.so.1 => /lib/x86_64-linux-gnu/libutil.so.1 (0x00007f9cefa1a000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f9cef66f000) libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f9cef46b000) /lib64/ld-linux-x86-64.so.2 (0x00007f9cf036d000)

On Feb 6, 2018, at 1:00 PM, shiro-ihara notifications@github.com wrote:

I didn't have a debugger, so I installed it. But my computer still send the message "Segmentation fault". The debug files seem to be not created. I have built boost libraries and set BOOSTDIR, which in the makefile, to reach it. Tclap is too . Is there something I should edit more?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/LLNL/paraDIS_lib/issues/1#issuecomment-363319656, or mute the thread https://github.com/notifications/unsubscribe-auth/AEmhaIzn_b38QBFQkIPuDtUB2emxkw33ks5tR-pkgaJpZM4R5CNd.

Rich Cook

It's not the time it takes to take the take that takes the time; it's the time it takes between the takes that takes the time. - Roy Scheider

lee218llnl commented 6 years ago

Please test with the latest commit https://github.com/LLNL/paraDIS_lib/commit/046d389797ec94e71750d93ed970313a0ef89168.