Add workaround for issue described in #6168.

[NlightNFotis]

This fixes the infinite unwinding by adding calls to do_indirect_call_and_rtti_removal()
before the contracts code invocation in goto-instrument/.

This is not yet finished, as it will need a test-case attached, but this is pushed so that
it can garner some initial comments regarding the approach.

• Each commit message has a non-empty body, explaining why the change was made.
• Methods or procedures I have added are documented, following the guidelines provided in CODING_STANDARD.md.
• The feature or user visible behaviour I have added or modified has been documented in the User Guide in doc/cprover-manual/
• Regression or unit tests are included, or existing tests cover the modified code (in this case I have detailed which ones those are in the commit message).
• My commit message includes data points confirming performance improvements (if claimed).
• My PR is restricted to a single feature or bugfix.
• White-space or formatting changes outside the feature-related changed lines are in commits of their own.

[NlightNFotis]

Hi @SaswatPadhi, would it be possible to have a look at this PR? It works-around the issue described in #6168 using Martin's suggestion.

However, I couldn't get it to work without introducing a call to apply_loop_contracts() in enforce_contracts(). I'm not sure about this step, but this is what was preventing the change in latest develop from working (it worked cleanly against the version described in the issue, cbmc-5.32.0).

Besides, I'm not sure if that addition is changing the semantics of your contract code. It seems that in 9cdca48#diff-1ff6abca57de19a68456666ed5c25ce09fcaae20fdc9c2b109cb60cb311c183aL1208 the call to apply_loop_contracts() is removed from a branch that could be exercised, without an equivalent call being reachable elsewhere, but that's just my understanding. Please let me know if there are issues with the current approach.

[SaswatPadhi]

Thanks for implementing the workaround, @NlightNFotis!

We separated the function contract flags and loop contracts flags recently, so --enforce-all-contracts only enforces function contracts (for all functions). Loop contracts don't have separate enforcement and replacement steps, just a single apply step using the flag --apply-loop-contracts.

So you could replace the --enforce-all-contracts flag with --apply-loop-contracts and undo the new call to apply_loop_contracts(). Thanks!

[codecov]

Codecov Report

Merging #6340 (de9e891) into develop (c06ccf1) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff            @@
##           develop    #6340   +/-   ##
========================================
  Coverage    75.89%   75.90%           
========================================
  Files         1515     1515           
  Lines       163971   163990   +19     
========================================
+ Hits        124452   124469   +17     
- Misses       39519    39521    +2     

Impacted Files	Coverage Δ
.../goto-instrument/goto_instrument_parse_options.cpp	68.29% <100.00%> (+0.03%)	⬆️
src/goto-instrument/contracts/contracts.cpp	93.87% <0.00%> (-0.92%)	⬇️
src/goto-instrument/contracts/assigns.cpp	96.66% <0.00%> (+4.71%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 14d35cc...de9e891. Read the comment docs.

[NlightNFotis]

Thanks @SaswatPadhi, the latest suggestions have now been implemented, and I moved the test to the regression/contracts folder.

This patch should now fix the issue you are facing in #6168 if you call goto-instrument --apply-loop-contracts.

[SaswatPadhi]

Awesome! Thank you so much, @NlightNFotis.

Looks like that resolved all of the CI errors too.