October 2002 - PPL-devel - BUGSENG Mailman

ppl/src Polyhedron.cc
by Roberto Bagnara 10 Oct '02

10 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: roberto(a)cs.unipr.it 2002-10-10 16:45:41 Modified files: src : Polyhedron.cc Log message: Do not use vector<bool>. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Polyhedron.cc.diff?cvsroo…

1 0

ppl/src Polyhedron.cc
by Elisa Ricci 10 Oct '02

10 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: ericci(a)cs.unipr.it 2002-10-10 11:20:44 Modified files: src : Polyhedron.cc Log message: Some improvements to the third technique of Polyhedron::BBRZ02_widening_assign(): - the test y.relation_with(x_g) == Poly_Gen_Relation::nothing() is done after testing if x_g is a ray; - corrected the test of the elements of the vector "considered"; - now we use a const reference x_g (for x.gen_sys[i]) to find how it evolved since a ray of "y" and a copy of "x_g" to built the new ray. Use "H79_con_sys" instead of "common_con_sys" in Polyhedron::BBRZ02_widening_assign() and in Polyhedron::H79_widening_assign(). Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Polyhedron.cc.diff?cvsroo…

1 0

ppl/src Polyhedron.cc Polyhedron.defs.hh
by Elisa Ricci 10 Oct '02

10 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: ericci(a)cs.unipr.it 2002-10-10 10:35:33 Modified files: src : Polyhedron.cc Polyhedron.defs.hh Log message: Improved the messages of the exceptions. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Polyhedron.cc.diff?cvsroo… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Polyhedron.defs.hh.diff?c…

1 0

ppl/tests addgenerators11.cc
by Roberto Bagnara 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: roberto(a)cs.unipr.it 2002-10-09 20:12:08 Modified files: tests : addgenerators11.cc Log message: Be quiet. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/addgenerators11.cc.diff…

1 0

Bug report
by P M Hill 09 Oct '02

09 Oct '02

Dear GNU developer, I have found the following problem in GNU Prolog. (Running Linux 2.4.2 and using GNU Prolog 1.2.11.) The bug is as follows: if the variable occurs before a choice point and is then bound to a non-variable by a predicate defined in an interface (to C++) after the choice point, then its binding is not undone on backtracking. More precisely, the bug occurs if: a variable is referred to before a choice point; there is a choice point; after the choice point, an C++ interface predicate binds the variable to a non-variable term; the procedure is then forced to backtrack by ensuring the first branch of the choice point fails. The program below has three tests. The first (test1) shows the bug. The other two (test2, test3) have no problem and by comparing them with test1 help indicate what features are needed to reproduce the bug. The tests test1 and test2 use ppl_new_Polyhedron_from_dimension/3. This is a predicate defined in the interface to the PPL (the Parma Polyhedra Library) which is written in C++. This predicate binds a variable P to a handle to the vector space with the specified number of dimensions (0) and topology (c). Before calling this predicate, P should be free. After a call to this predicate, P should be bound to a term of the form $address(N,M), where N and M are numbers. The other PPL predicates (ppl_initilize/0, ppl_finalize/0, ppl_delete_Polyhedron/1) used in the tests are relevant for using the PPL but are not directly related to the problem being reported. To run the tests you need both this little program and the PPL. Note that the PPL with documentation is available at http://www.cs.unipr.it/ppl/ I am hoping that, even if you do not download and install the PPL, you may be able to find the cause of the bug from the description here which seems to be due to the interraction between using the interface and backtracking. Also, if it helps, I can perform for you any further tests you suggest. ------------------------------------------------------------------ % The program gnu_bug.pl to show a bug in GNU Prolog 1.2.11 /* The first test shows that if the variable occurs before the choice point and is then bound to a handle for a polyhedron after the choice point, then its binding is not undone on backtracking. The second test shows that if the variable does not occur before the choice point, then the binding is undone ok. The third test shows that if it does not use the ppl interface, then the binding is undone ok. */ test1 :- nl, write('test1 has the 1st occurrence of P before the choice point'), nl, ppl_initialize, nl, write('Before the choice point:'), write(' N = '), write(N), write(' P = '), write(P), nl, nl, write('After the choice point:'), try(N), write(' N = '), write(N), write(' P = '), write(P), ppl_new_Polyhedron_from_dimension(c, 0, P), cleanup(P), N=2, nl, ppl_delete_Polyhedron(P), !, ppl_finalize. test2 :- nl, write('test2 has the 1st occurrence of P after the choice point'), nl, ppl_initialize, %% nl, write('Before the choice point:'), %% write(' N = '), write(N), write(' P = '), write(P), nl, nl, write('After the choice point:'), try(N), write(' N = '), write(N), write(' P = '), write(P), ppl_new_Polyhedron_from_dimension(c, 0, P), cleanup(P), N=2, nl, ppl_delete_Polyhedron(P), !, ppl_finalize. test3 :- nl, write('test3 has the same structure as test1 without the PPL interface'), nl, ppl_initialize, nl, write('Before the choice point:'), write(' N = '), write(N), write(' P = '), write(P), nl, nl, write('After the choice point:'), try(N), write(' N = '), write(N), write(' P = '), write(P), P = 99, N=2, nl, !, ppl_finalize. try(1). try(2). % To prevent leaks: cleanup(_Polyhedron). cleanup(Polyhedron) :- ppl_delete_Polyhedron(Polyhedron), fail. --------------------------------------------------------------------------- Below I provide the output from the tests using GNU Prolog, and, for comparison, the output from the same tests using SWI Prolog. $ /bin/sh /.../libtool --mode=execute -dlopen /.../libppl.la ./ppl_gprolog GNU Prolog 1.2.11 By Daniel Diaz Copyright (C) 1999-2002 Daniel Diaz | ?- [gnu_bug]. compiling /.../gnu_bug.pl for byte code... /.../gnu_bug.pl compiled, 58 lines read - 4931 bytes written, 424 ms yes | ?- test1. test1 has the 1st occurrence of P before the choice point Before the choice point: N = _273 P = _274 After the choice point: N = 1 P = _274 N = 2 P = $address(32008,2093) yes | ?- test2. test2 has the 1st occurrence of P after the choice point After the choice point: N = 1 P = _273 N = 2 P = _273 yes | ?- test3. test3 has the same structure as test1 without the PPL interface Before the choice point: N = _273 P = _274 After the choice point: N = 1 P = _274 N = 2 P = _274 yes | ?- halt. $ cd ../SWI $ /bin/sh /.../libtool --mode=execute -dlopen /.../src/libppl.la ./ppl_pl Welcome to SWI-Prolog (Version 5.0.2) Copyright (c) 1990-2002 University of Amsterdam. SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software, and you are welcome to redistribute it under certain conditions. Please visit http://www.swi-prolog.org for details. For help, use ?- help(Topic). or ?- apropos(Word). ?- [gnu_bug]. % gnu_bug compiled 0.00 sec, 3,080 bytes Yes ?- test1. test1 has the 1st occurrence of P before the choice point Before the choice point: N = _L154 P = _L155 After the choice point: N = 1 P = _L155 N = 2 P = _L155 Yes ?- test2. test2 has the 1st occurrence of P after the choice point After the choice point: N = 1 P = _L155 N = 2 P = _L155 Yes ?- test3. test3 has the same structure as test1 without the PPL interface Before the choice point: N = _L154 P = _L155 After the choice point: N = 1 P = _L155 N = 2 P = _L155 Yes ?- halt. $ */ Best regards, Patricia Hill

1 0

ppl/tests bbrz02widening9.cc bbrz02widening13. ...
by Elisa Ricci 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: ericci(a)cs.unipr.it 2002-10-09 17:59:59 Modified files: tests : bbrz02widening9.cc bbrz02widening13.cc bbrz02widening14.cc bbrz02widening15.cc Log message: Added the "known_result" to verify if the resulting polyhedron is equal to the polyhedron that we want to have. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/bbrz02widening9.cc.diff… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/bbrz02widening13.cc.dif… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/bbrz02widening14.cc.dif… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/bbrz02widening15.cc.dif…

1 0

ppl/interfaces/Prolog clpq2.pl
by Patricia Hill 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: hill(a)cs.unipr.it 2002-10-09 14:54:45 Modified files: interfaces/Prolog: clpq2.pl Log message: Problem due to bug in GNU Prolog now avoided by delaying as much as possible any reference to variables that are intended to be bound to polyhedral handles. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/interfaces/Prolog/clpq2.pl.di…

1 0

ppl/src algorithms.hh
by Enea Zaffanella 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2002-10-09 11:28:39 Modified files: src : algorithms.hh Log message: poly_hull_assign_if_exact(p,q) now return true if the poly-hull is exact. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/algorithms.hh.diff?cvsroo…

1 0

ppl/src Constraint.defs.hh Polyhedron.defs.hh
by Enea Zaffanella 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2002-10-09 10:55:31 Modified files: src : Constraint.defs.hh Polyhedron.defs.hh Log message: Escaped a few ">" and "<" occurring inside HTML tags. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Constraint.defs.hh.diff?c… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Polyhedron.defs.hh.diff?c…

1 0

ppl/doc definitions.dox
by Enea Zaffanella 09 Oct '02

09 Oct '02

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2002-10-09 10:28:59 Modified files: doc : definitions.dox Log message: "<" and ">" inside HTMK tags have to be escaped by "\". Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/definitions.dox.diff?cvsr…

1 0