Bug 264, which this workaround was for, has been fixed in other changes.
compiler/mode_errors.m:
Remove the temporary error.
compiler/modecheck_unify.m:
Don't check this error condition.
doc/reference_manual.texi:
Document the constraint that actually applies, namely that functions
that do not match the default mode do not match Mercury's 'ground'
inst.
Clarify that "no higher-order inst information" means no *explicit*
information. Given that there is a default, this is not the same as
saying there are no logical consequences.
tests/invalid/default_ho_inst.err_exp:
tests/invalid/default_ho_inst_2.err_exp:
Update expected output.
tests/valid/Mmakefile:
tests/valid/ho_func_call_2.m:
New test case. Function modes that are not the same as the default,
but that match it, can still safely be treated as ground.
The code that takes the function out of the term can use the function
as if it had the default signature. If it doesn't, this can lead to
This fixes Mantis bug 264.
compiler/mode_errors.m:
Provide a way to record such errors, and to print error messages for them.
compiler/modecheck_unify.m:
Look for such errors when constructing terms.
browser/declarative_user.m:
Don't put a function with a nondefault signature (it used a subtype)
into a term. Enforce the subtype with a runtime assertion instead.
doc/reference_manual.texi:
Document the new limitation.
tests/invalid/default_ho_inst.{m,err_exp}:
The test case for Mantis bug 264.
tests/invalid/Mmakefile:
Enable the new test case.
tests/valid/Mmakefile:
Disable the old ho_func_call test case, since it RELIES on putting
functions with nondefault modes into terms. It also uses I/O states
in ui modes.
