While using gprolog I often have exceptions without any kind of line numbers or context like this one:
uncaught exception: error(instantiation_error,(is)/2)
Without any kind of context. I know I can do a trace
but it would take very long to debug it with trace
since I need to execute a lot of things before getting to the place where the error occur.
Any idea on how to have this stacktrace ? Or a dynamic trace
/ notrace
?
EDIT: Or just automate the printing of the whole trace
output.
@gusbro's answer (
s(X)
) shows you how you somewhat solve this with GNU's debugger. However, if you cannot afford to see all the printing going on, or it is much too slow, you might consider the following "debugger" available aslibrary(debug)
in Scryer.I personally do not use debuggers offered by Prolog systems for the simple reason that most of them print much too much, are often buggy themselves, and have their own specific ever changing conventions, I can't afford to learn.
To use it, simply add
$-
,$
, or@
in front of a specific goal.$-
means: only signal exceptions going through this goal$
additionally show call and exit@
assures that there is at least one answer, and if not, it is reported and an exception is thrown.Use above annotations sparingly!
*
removes the goal. This is for generalizing a program doing program modification/program-slicing in a pure monotonic program. You need this in case of unexpected failures. For examples how to use it, see the following answers/debugging sessions 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13. More._/*term*/
replaces a term by an anonymous variable. This generalizes a program even further than*
alone. Example sessions: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, In this manner you can reduce the information you watch significantly.In other systems supporting
meta_predicate
directives like SICStus, YAP and SWI, add in front the following directive:And then, for cases of (suspected) non-termination or instantiation errors, consider to use a failure-slice.