PetscAttachDebuggerErrorHandler#

Error handler that attaches a debugger to a running process when an error is detected. This routine is useful for examining variables, etc.

Synopsis#

#include "petscsys.h"   
PetscErrorCode PetscAttachDebuggerErrorHandler(MPI_Comm comm, int line, const char *fun, const char *file, PetscErrorCode num, PetscErrorType p, const char *mess, void *ctx)

Not Collective, No Fortran Support

Input Parameters#

comm - communicator over which error occurred
line - the line number of the error (usually indicated by __LINE__ in the calling routine)
fun - the function name of the calling routine
file - the file in which the error was detected (usually indicated by __FILE__ in the calling routine)
mess - an error text string, usually just printed to the screen
num - the generic error number
p - PETSC_ERROR_INITIAL if error just detected, otherwise PETSC_ERROR_REPEAT
ctx - error handler context

Notes#

By default the GNU debugger, gdb, is used except on macOS where lldb is used. Alternatives are cuda-gdb, lldb, dbx and xxgdb,xldb (on IBM rs6000), xdb (on HP-UX).

Most users need not directly employ this routine and the other error handlers, but can instead use the simplified interface SETERR, which has the calling sequence

  SETERRQ(PETSC_COMM_SELF, number, p, message)

Use PetscPushErrorHandler() to set the desired error handler. The currently available PETSc error handlers are

    PetscTraceBackErrorHandler()
    PetscAttachDebuggerErrorHandler()
    PetscAbortErrorHandler()

or you may write your own.

Developer Note#

This routine calls abort instead of returning because if it returned then MPI_Abort() would get called which can generate an exception causing the debugger to be attached again in a cycle.

Level#

developer

Location#

src/sys/error/adebug.c

Index of all Sys routines
Table of Contents for all manual pages
Index of all manual pages