Lab Course SoSe2025 - Pentagon Domain #1740

feniup · 2025-05-07T11:08:48Z

Lab Course: Static Analysis with Goblint

In this lab course we try to actively contribute to the Goblint codebase by integrating new abstract domain.

In the current iteration, the focus is on the development of the pentagon domain (Pntg), which combines interval analysis with symbolic relational constraints of the form $x < y$. This hybrid approach offers greater precision than traditional interval analysis while maintaining lower complexity compared to more expressive domains like the octagon domain.

…nagment or Apron stuff.

….leq

michael-schwarz · 2025-07-17T08:39:08Z

src/cdomains/pentagon/zExt.ml

+  let hash (z: t) = 
+    match z with
+    | PosInfty -> failwith "ZExt.pow: TODO" 
+    | NegInfty -> failwith "ZExt.pow: TODO"
+    | Arb(z) -> Z.hash z;;


I am bit surprised your analysis works at all, given it will fail if any PosInfinity or NegInfinity appears in an abstract value. What precisely is your plan for how you use them?

michael-schwarz · 2025-07-17T08:42:11Z

tests/unit/dune

+; See: https://github.com/ocaml/dune/issues/4383#issuecomment-805107435.
+; TODO: Remove workaround with dune 3.0, where this should get fixed.


I don't think the entire comment needs to be repeated.

michael-schwarz · 2025-07-17T08:47:13Z

tests/unit/cdomains/pentagonDomain/pentagonTest.no-apron.ml

+(* Note: 
+   There are currently no tests independent of apron.
+*)


If they are not independent of apron, why are they called No-Apron?

michael-schwarz · 2025-07-17T08:47:30Z

tests/unit/cdomains/pentagonDomain/pentagonTest.apron.ml

+  ) |>
+  assert_equal expected
+
+let noop _ = assert_bool "" true


michael-schwarz · 2025-07-17T08:59:06Z

src/cdomains/pentagon/boxes.ml

+  let narrow (i1: t) (i2: t) = 
+    meet i1 i2


Could you elaborate on why setting narrow to meet is ok here, i.e., why the domain does not have (quasi-)infinite (i.e., of length 2^64 or something) descending chains?

michael-schwarz · 2025-07-17T09:03:35Z

src/cdomains/pentagon/intv.ml

+  let narrow (i1: t) (i2: t) = 
+    inter i1 i2


Same question as above for narrow.

michael-schwarz · 2025-07-17T09:05:32Z

src/cdomains/pentagon/intv.ml

+*)
+module Intv =
+struct
+  include ZExtOps


This is almost a lattice except you have named (some) operations differently? Any reasons for that?

michael-schwarz · 2025-07-17T09:06:55Z

src/cdomains/pentagon/intv.ml

+
+  type t = (ZExt.t * ZExt.t) [@@deriving eq, hash, ord]
+
+  let to_string ((lb, ub): t) = Printf.sprintf "[%s,%s]" (StringUtils.int32_bound_str lb) (StringUtils.int32_bound_str ub)


Why Int32?

…l evaluation

… some explaining comments; Minor performance stuff;

Work

src/cdomains/apron/pentagonDomain.apron.ml

+          | ZExt.PosInfty -> failwith "This should not happen :)"
+          | ZExt.NegInfty -> (
+              if Z.abs coeff = Z.one then
+                if List.length l >= 2 then


Work

DrMichaelPetter · 2025-07-25T11:53:23Z

src/cdomains/apron/pentagonDomain.apron.ml

+              "Failed to convert cil expression: exception %s"
+              (SharedFunctions.show_unsupported_cilExp exn),


This is where at least on my computer the memory overflow happens; can you do without the call to show_unsupported_cilExp ?

I guess we can remove this call. iirc we only put this in for debugging purposes.

PS: How did you recognize it as the cause?

I started goblint/pentagon with the command from an SV-Comp OUT OF MEMORY

Waited for the memory usage to build up, without progress in the eval counter

pressed ctr-c

inspected the stacktrace for the uppermost occurance of a line from the pentagon implementation

Is this consistent, i.e., does redoing those steps always result in that spot? It might just be a coincidence that the Ctrl-C interrupt happened right at that call.

Nevertheless, I will do some experiments later today and check it out.

It seems to be d_plainexp in the CIL Pretty printer. We've had similar problems with recursive printers in the past.

AHHH now I can see it.
I miss interpreted the calls to the Pretty module and just "waved them off", as I thought they were needed to produce the output when specifying ./goblint -v.

You are right. d_plainexp starts a long chain of calls.

I just simply removed the lines for a simple test and it seems to make the cut. Wow, I really did not anticipate that easy of a fix! That's great 😄

DrMichaelPetter · 2025-07-25T12:26:21Z

src/cdomains/apron/pentagonDomain.apron.ml

+    let (str, res) = 
+      match Convert.tcons1_of_cil_exp ask t t.env e negate no_ov with 
+      | exception Convert.Unsupported_CilExp exn -> (
+          (
+            Printf.sprintf 
+              "Failed to convert cil expression: exception %s"
+              (SharedFunctions.show_unsupported_cilExp exn),
+            t
+          )
+        )
+      | tcons -> (
+          (
+            StringUtils.string_of_tcons1 tcons,
+            assert_constraint ask t tcons negate no_ov
+          )
+        )
+    in


This should go into the tracing part

For the PR; Created an initial file that differs;

05b3577

michael-schwarz added student-job relational Relational analyses (Apron, affeq, lin2var) feature labels May 7, 2025

Added new empty files for new pntg analysis and domain

b76013a

sim642 added the in progress label May 8, 2025

Philip Feniuk and others added 24 commits May 13, 2025 16:49

First structure of the new pentagon domain

9210b70

Added top and bot functions to both modules

69f7978

Type annotations and Inequlities module

2fdf9a3

SUB leq

599e38c

leq and join

1bcdeee

rewritten leq and join to adhere to Apron Environment type

a9521a5

Implemented meet and widen, fixed leq in SUB

3297211

Implementations for SUB.{top, is_top, narrow}

5915bfb

Added first naive implementation of intervals without including VarMa…

45eecfa

…nagment or Apron stuff.

Split domain operations and created methods for single intervals.

93d738b

Unit test setup

94dc528

Unfinished stuff; Fixed SUB functions to work with new environment stuff

0aed838

Added first test cases

1dde07c

Fixed pentagon leq function

e9d3376

Merge branch 'work' of github.com:feniup/analyzer-pentagon into work

938a17c

removed unnecessary comment

9a8452b

added narrow and meet

c33e8ed

pentagon join implementation

4f781fc

Merge branch 'work' of github.com:feniup/analyzer-pentagon into work

50faefd

added test cases for sub domain

e474403

Unit tests and some fixes

f59488b

Added necessary dune config for pentagon unit testing

bd7996e

Fixed some unit tests and the sub functionality of SUB.is_bot and SUB…

d786d42

….leq

i broke everything on purpose

5f415cf

Printing texpr and intv for bound_texpr

d98722d

michael-schwarz reviewed Jul 17, 2025

View reviewed changes

tests/unit/cdomains/pentagonDomain/pentagonTest.apron.ml

) |>

assert_equal expected

let noop _ = assert_bool "" true

Copy link

Member

michael-schwarz Jul 17, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

?

michael-schwarz reviewed Jul 17, 2025

View reviewed changes

Alexander Mazur and others added 8 commits July 17, 2025 13:52

work in progress

3e0821a

added consistency check between linearized and non-linearized interva…

8e8bed2

…l evaluation

adapted narrowing for intb

538c94a

adapted boses

51e1f83

Fixed conversion from scalar to zext; Fixed narrowing for Pntg; Added…

040f5da

… some explaining comments; Minor performance stuff;

Add gcc to creduce script

1a8d3ce

Remove narrow for Subs; Timing.wrap for dim_add and dim_remove;

62fc115

Merge pull request #6 from feniup/work

21a3ce6

Work

github-advanced-security bot found potential problems Jul 23, 2025

View reviewed changes

Philip Feniuk and others added 5 commits July 23, 2025 15:30

Merge remote-tracking branch 'origin/master' into work

29d32e8

Added sub usage in bound_texpr and assign_texpr

973e7a9

Adapted regression test because of added functionality

f43c458

Merge pull request #7 from feniup/work

25e3913

Work

Merge remote-tracking branch 'upstream/master'

9e59acc

DrMichaelPetter reviewed Jul 25, 2025

View reviewed changes

Philip Feniuk and others added 5 commits July 25, 2025 18:27

Fixed pentagon memory issue

f73bce5

Reverted printing in Lin2Var assert_constraint

9c0fd4a

Remove double semicolons

e09d123

Renamed eval_monoms_to_intv_infty_list

4006c89

sign bug fix

fa828c4

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Lab Course SoSe2025 - Pentagon Domain #1740

Lab Course SoSe2025 - Pentagon Domain #1740

feniup commented May 7, 2025 •

edited

Loading

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025 •

edited

Loading

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

michael-schwarz Jul 17, 2025

Uh oh!

Check warning

DrMichaelPetter Jul 25, 2025

Uh oh!

feniup Jul 25, 2025

Uh oh!

DrMichaelPetter Jul 25, 2025

Uh oh!

feniup Jul 25, 2025

Uh oh!

michael-schwarz Jul 25, 2025

Uh oh!

feniup Jul 25, 2025

Uh oh!

DrMichaelPetter Jul 25, 2025

Uh oh!

Uh oh!

		; See: https://github.com/ocaml/dune/issues/4383#issuecomment-805107435.
		; TODO: Remove workaround with dune 3.0, where this should get fixed.


		type t = (ZExt.t * ZExt.t) [@@deriving eq, hash, ord]

		let to_string ((lb, ub): t) = Printf.sprintf "[%s,%s]" (StringUtils.int32_bound_str lb) (StringUtils.int32_bound_str ub)

		"Failed to convert cil expression: exception %s"
		(SharedFunctions.show_unsupported_cilExp exn),

Lab Course SoSe2025 - Pentagon Domain #1740

Are you sure you want to change the base?

Lab Course SoSe2025 - Pentagon Domain #1740

Conversation

feniup commented May 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Lab Course: Static Analysis with Goblint

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

michael-schwarz Jul 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Check warning

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

feniup commented May 7, 2025 •

edited

Loading

michael-schwarz Jul 17, 2025 •

edited

Loading