compiler/add_mode.m:
Improve diagnostics for circular insts and modes in several ways.
- Generate one error message per cycle, not one error message per inst.
In the case of cycles containing more than one inst or mode, this
reduces clutter.
- Unlike the old code, we do print a context for every inst or mode
in a cycle, rather than just for the first.
- Avoid module qualifying local inst and mode names unless that
qualification actually avoids ambiguity.
- Use more expansive wording that actually explains *why* a circular
inst or mode is a problem.
tests/invalid/circ_inst5.m:
Extend this test case to test corner cases of the new code.
tests/invalid/circ_inst.err_exp:
tests/invalid/circ_inst2.err_exp:
tests/invalid/circ_inst3.err_exp:
tests/invalid/circ_inst4.err_exp:
tests/invalid/circ_inst5.err_exp:
tests/invalid/circ_mode.err_exp:
tests/invalid/circ_mode2.err_exp:
tests/invalid/circ_mode3.err_exp:
tests/invalid/circ_mode4.err_exp:
Expect the updated error messages.
It allowed circular insts to be added to the HLDS undetected, which meant
a later pass got into an infinite loop.
compiler/make_hlds_passes.m:
Check for circularities in inst definitions only *after* all insts
have been added to the HLDS, because until then, a circularity check
may fail not because of the absence of a circularity in the inst
definitions in the program, but because of the absence of a circularity
in the inst definitions *added to the HLDS so far*.
Do likewise for mode definitions.
compiler/add_mode.m:
To make the above possible, split the predicates that do
circularity checks on insts and modes from the predicates that
add insts and modes to the HLDS.
tests/invalid/circ_inst5.m:
Reorder inst definitions to expose the infinite loop bug in compilers
without this diff.
tests/invalid/circ_inst*.err_exp:
tests/invalid/circ_mode*.err_exp:
In the case of mutually recursive pairs of insts or modes, expect
error messages for both definitions, not just the later definition.
tests/invalid/*.{m,err_exp}:
tests/misc_tests/*.m:
tests/mmc_make/*.m:
tests/par_conj/*.m:
tests/purity/*.m:
tests/stm/*.m:
tests/string_format/*.m:
tests/structure_reuse/*.m:
tests/submodules/*.m:
tests/tabling/*.m:
tests/term/*.m:
tests/trailing/*.m:
tests/typeclasses/*.m:
tests/valid/*.m:
tests/warnings/*.{m,exp}:
Make these tests use four-space indentation, and ensure that
each module is imported on its own line. (I intend to use the latter
to figure out which subdirectories' tests can be executed in parallel.)
These changes usually move code to different lines. For the tests
that check compiler error messages, expect the new line numbers.
browser/cterm.m:
browser/tree234_cc.m:
Import only one module per line.
tests/hard_coded/boyer.m:
Fix something I missed.
Estimated hours taken: 0.5
Branches: main, release
Fix another problem with circularly defined insts.
compiler/make_hlds.m:
When checking for circularly defined insts, substitute the correct
arguments for user defined insts. Use unrelated insts as the
inital argument values (in this case, use not_reached for all of
them).
tests/invalid/Mmakefile:
Add some new tests.
tests/invalid/circ_inst5.err_exp:
tests/invalid/circ_inst5.m:
A test case where looping occurs via an argument substitution.
Before this change, attempting to compile this module would result
in a loop.
tests/invalid/circ_type5.err_exp:
tests/invalid/circ_type5.m:
A similar test case for equivalence types. The algorithm to expand
equivalence types already handled this case correctly, but it was
untested.
There is no corresponding test case for modes, since this kind of
error cannot occur in mode declarations.
