We have a very simple requirement of calling a couple of native functions from Java. We are using JNA for making those native calls.
Edits: We don't have any custom native code. We are making calls to Linux Kernel C library functions.
We are getting very weird memory corruption errors like
The program even hangs up some times. These errors are intermittent.
Some standard examples/docs around using structures in JNA calls would be helpful.
This is our library wrapper having native functions:
These are the native models which map to structures in C:
And this is how we are accessing the library functions:
val timeVal = new NTPTimeVal()
TimeLibrary.ntp_gettimex(timeVal)
println(timeVal.tai)
You can refer to the TimeServiceImpl.scala for more clarity.
Could someone tell us what exactly are we doing wrong?
There are some reserved fields in ntptimeval and related structures:
struct ntptimeval
{
struct timeval time; /* current time (ro) */
long int maxerror; /* maximum error (us) (ro) */
long int esterror; /* estimated error (us) (ro) */
long int tai; /* TAI offset (ro) */
long int __glibc_reserved1;
long int __glibc_reserved2;
long int __glibc_reserved3;
long int __glibc_reserved4;
};
Which you don't have in your code:
public class NTPTimeVal extends Structure {
public TimeVal time; /* Current time */
public Long maxerror; /* Maximum error */
public Long esterror;
public int tai;
}
If those reserved fields happen to be used in your glibc version, that can explain the heap corruption.
I would also carefully examine the data you get back. If some fields contain strange values, it may mean field size/alignment problem, which can also signal the structure being shorter than it needs to be.