invalid pointer error

wl · 03-22-2017, 10:27 AM

OS: Linux redhat 6.8
Software version using: Open R 3.2.2
Error about: error while running job in batch mode using command "R CMD BATCH test.R"

I am using open source software called "open R". while running one R script "test.R" in batch mode - receiving below error about invalid pointer,

*** glibc detected *** /usr/lib/rstudio-server/bin/rsession: free(): invalid pointer: 0x00000000025d3d20 ***

I believe this is OS level error as same script running fine from one of R tool,

any suggestion?

rtmistler · 03-22-2017, 11:24 AM

Your system could have run out of memory and the logic of the code didn't pay attention to an allocation return error, but eventually when it went to free memory, it found that the pointer was invalid.

wl · 03-22-2017, 12:00 PM

Thanks for your note,

infect I was monitoring server activities and there was no sign of max memory or max cpu utilization or anything else - server was normal at all time,,,

rtmistler · 03-22-2017, 12:14 PM

Do you have the source for Open R 3.2.2? Seems as if you'll need to debug the code.

Where exactly did you see this error report? In the system log, on the console?

Is rstudio related to open R? The error has occurred in rstudio by the way.

wl · 03-22-2017, 02:29 PM

rstudio (gui) is a web R tool,

We have same code running fine in rstudio - error we are getting is from command line when we try running job in batch mode,

rtmistler · 03-22-2017, 02:34 PM

Well besides trial and error to discern the similarities and differences between method A and B as to how you run this and learn something from that, the only other option for me would be to build the code with a debug flag and run it on a debugger.

pan64 · 03-22-2017, 02:39 PM

I do not know what do you mean by OS level error. I would rather say this is a problem with the binary executable itself - or a library (.so) used by that.
Remember scripts themselves are not running, they are always executed by another app, usually an interpreter or similar.
debugging can be a good idea to catch it, or probably there is a coredump to analyse. Maybe you can construct a bug report too, if it was reproducible.