Dynamic predicates are sometimes confused by unrelated libraries

[triska]

To reproduce this issue, please create file.pl comprising:

:- use_module(library(clpz)).

clpz:monotonic.

Consulting this file on the toplevel works exactly as expected:

$ scryer-prolog -f
?- [file].
   true.

When I then keep the process running and change the file to read:

:- use_module(library(lists)).
:- use_module(library(clpz)).
:- use_module(library(dcgs)).

% clpz:monotonic.

i.e., I have added 2 additional libraries, lists and dcgs, respectively before and after clpz, and commented out the fact clpz:monotonic, I get in continuation of the earlier toplevel interaction:

?- [file].
   true.
?- X #> 3.
   clpz:(X in 4..sup).
?- X #> 3.
   clpz:(X in 4..sup).
?- X #> 3.
   clpz:(X in 4..sup).
?- [file].
   true.
?- X #> 3.
   error(existence_error(procedure,monotonic/0),monotonic/0).
   error(existence_error(procedure,monotonic/0),monotonic/0).

Note the highly unexpected behaviour: Everything works exactly as expected when I (re-)consult the file once. Only after I re-consult it a second time do I get an unexpected existence error when posting X #> 3.

clpz:monotonic is declared as a dynamic predicate by library(clpz). I therefore expect it to fail if it is not defined.

Where exactly it is defined should have no influence on its dynamic property, and loading completely unrelated libraries should definitely have no influence at all on this property! That I get an existence error twice is even more unexpected!

[triska]

This issue persists, I only need to load file.pl a third time:

?- [file]. % with the original file
   true.   % now make the changes mentioned above
?- [file]. % load the changed file
   true.
?- X #> 3.
   clpz:(X in 4..sup). % it appears to work
?- [file]. % load the changed file again
   true.
?- X #> 3.
   error(existence_error(procedure,monotonic/0),monotonic/0).
   error(existence_error(procedure,monotonic/0),monotonic/0).

[triska]

Well done, thank you a lot!