Project

General

Profile

Bug #1717

CRASH: equality test to a FUNCTION

Added by John Abbott over 1 year ago. Updated over 1 year ago.

Status:
Closed
Priority:
Urgent
Assignee:
John Abbott
Category:
bug
Target version:
CoCoA-5.4.2
Start date:
01 Dec 2022
Due date:
% Done:

100%

Estimated time:
1.11 h
Spent time:
1.15 h

Description

Julian Danner reported the following crash via email:

/**/  use R::=QQ[x]; phi := CoeffEmbeddingHom(R); phi = func() endfunc;
CoCoAInterpreter: Interpreter.C:1581: boost::intrusive_ptr<CoCoA::InterpreterNS::RightValue> CoCoA::InterpreterNS::RuntimeEnvironment::binaryOperatorDispatch(boost::intrusive_ptr<const CoCoA::InterpreterNS::RightValue>, boost::intrusive_ptr<const CoCoA::InterpreterNS::RightValue>, boost::intrusive_ptr<CoCoA::InterpreterNS::RightValue> (* (*)[24])(CoCoA::InterpreterNS::RuntimeEnvironment*, boost::intrusive_ptr<const CoCoA::InterpreterNS::RightValue>, boost::intrusive_ptr<const CoCoA::InterpreterNS::RightValue>, const CoCoA::LexerNS::CharPointer&, const CoCoA::LexerNS::CharPointer&), const CoCoA::LexerNS::CharPointer&, const CoCoA::LexerNS::CharPointer&): Assertion `leftOp->getType()->dispatchIndex == rightIndex' failed.

Process cocoa5 aborted (core dumped)

Ouch! It was originally a typo := mistyped as =

History

#1 Updated by John Abbott over 1 year ago

Most likely the bug is in intrusive_ptr<RightValue> RuntimeEnvironment::binaryOperatorDispatch
around line 1520 in Interpreter.C

#2 Updated by John Abbott over 1 year ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 10

The relevant assert is at Interpreter.C:1581

A slightly shorter failing case is:

func() endfunc = IdentityHom(ZZ);

In many cases CoCoA detects that FUNCTION cannot be compared to the other type, but RINGHOM is not so reported.
Odd!

#3 Updated by John Abbott over 1 year ago

  • Status changed from In Progress to Resolved
  • Assignee set to John Abbott
  • % Done changed from 10 to 70

Here is what I found.
The decision about which types may be compared with = is determined by opEqualMap.
This had a fn pointer to opEqual_Function_Function which just threw an exception ("Functions cannot be compared").

However if there is no entry in OpEqualMap for (FUNCTION,FUNCTION) then essentially the same error mesg is produced
("I don't know how to ...").

Note that RINGHOM is regarded as a subtype of FUNCTION via function bool TYPE::isProperSubtypeOf
around Interpreter.C:1467. I am loath to change that function for fear that it might break something
e.g. when using a RINGHOM in a "function call context".

Instead I have removed the (FUNCTION,FUNCTION) entry from opEqualMap and from opNotEqualMap.

All CoCoA-5 tests pass.

Now CoCoA-5 reacts as follows:

/**/ func() endfunc = IdentityHom(ZZ);
--> ERROR: I don't know how to evaluate operator = between FUNCTION and RINGHOM
--> func() endfunc = IdentityHom(ZZ);
-->                ^

I think this resolves the issue. Comments? Criticisms?

#4 Updated by John Abbott over 1 year ago

  • Status changed from Resolved to Closed
  • % Done changed from 70 to 100
  • Estimated time set to 1.11 h

We are satisfied that this is a good resolution.

Also available in: Atom PDF