compiler/prog_item.m:
We used to record information about include declarations
in parse_tree_int[012] in two forms:
- as a pair of maps from module names to contexts (one each for
includes in the interface and implementation sections), and
- as a single map from module names to an include_module_info, which
recorded the section of its appearance along with its context.
The second of these data structures is derived from the first,
in a process that can result in the generation of diagnostic messages.
In the absence of any issues reported by these diagnostics, the two forms
contain the same information.
Avoid this redundancy by keeping only the second form in the parse trees
of .int0, .int and .int2 files. (.int3 files cannot contain include_module
declarations.)
Since .int2 files may contain include_module declarations only in
the interface section, change the representation of the second form
to a type that expresses this invariant: int_include_module_map,
which is a subtype of the existing type include_module_map.
compiler/comp_unit_interface.m:
compiler/convert_parse_tree.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/parse_tree_out.m:
compiler/recompilation.check.m:
compiler/recompilation.version.m:
Conform to the change above.
compiler/item_util.m:
Add a utility predicate for use by new code above.
compiler/prog_item.m:
We used to record information about import and use declarations
in parse_tree_int[012] in two forms:
- as a quartet of maps from module names to contexts (one each for
int imports, int uses, imp imports and imp uses), and
- as a single map from module names to a section_import_and_or_use,
which recorded the section and kind (import or use) of its appearance
along with its one context, except for the case of modules that have
an use_module declaration in the interface section and an import_module
declaration in the implementation section.
The second of these data structures is derived from the first,
in a process that can result in the generation of diagnostic messages.
In the absence of any issues reported by these diagnostics, the two forms
contain the same information.
Avoid this redundancy by keeping only the second form in the parse trees
of .int0, .int and .int2 files. (For .int3 files, which can contain
only import_modules, and only in the interface section, this redundancy
has not been present even before now.)
Since .int and .int2 files may contain only use_module declarations
and not import_module declarations, change the representation of the
second form to a type that expresses this invariant: the new type
section_use_map, which is a subtype of the existing type
section_import_and_or_use_map.
For .int2 files, we could use an even tighter type right now, but
a fix for Mantis bug #563 would have to undo such a change, so
don't bother.
compiler/comp_unit_interface.m:
Delete the code that used to construct the first form above
for these interface file kinds. Conform to the changes above.
compiler/convert_parse_tree.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/parse_tree_out.m:
compiler/recompilation.check.m:
compiler/recompilation.version.m:
Conform to the changes above.
compiler/item_util.m:
Add new, specialized versions of existing utility predicates
to make that conformance possible.
compiler/prog_item.m:
The ptiN_import_use_map fields in the representations of .int0, .int
and .int2 files had the same type as the ptms_import_use_map field
in the parse trees of .m files, which is where they were derived from.
However, while the ptms_import_use_map field needs to be able to represent
implicit imports, the parse trees of .int0, .int and .int2 files
should never include any implicit imports, and in fact any implicit
imports in these fields were already ignored.
Encode the invariant that interface files never include implicit imports
in the types of these fields.
compiler/comp_unit_interface.m:
Discard the implicit part of the source file's import_and_or_use_map
when computing the contents of .int0, .int and .int2 files.
compiler/item_util.m:
Provide the facilities used by the updated code in the modules above.
compiler/convert_parse_tree.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
Conform to the changes above.
compiler/mercury_compile_main.m:
compiler/mercury_compile_front_end.m:
compiler/mercury_compile_llds_back_end.m:
compiler/mercury_compile_make_hlds.m:
compiler/mercury_compile_middle_passes.m:
compiler/mercury_compile_mlds_back_end.m:
Pass progress and error streams explicitly in these top modules
of the compiler. Use "XXX STREAM" to mark places where we could switch
from using stderr for both the progress and error streams to using
module-specific files as the progress and/or error streams.
compiler/passes_aux.m:
Add a "maybe_" prefix to the names of the predicates that print progress
messages at the appropriate verbosity levels, as their printing of those
messages is conditional.
Provide versions of those predicates that take explicitly specified
streams to write to, and mark the versions that write to the current
output stream as obsolete.
The predicate that wrote progress messages for procedures
used to have two versions, one taking a pred_proc_id, and one taking
a pred_id/proc_id pair. Delete the latter, because the arity difference
that differentiated the two versions is now needed for the difference
between supplying and not supplying an explicit stream.
compiler/file_util.m:
compiler/hlds_error_util.m:
compiler/write_error_spec.m:
Delete several predicates that wrote to the current output stream,
since all their callers now use the versions that specify an explicit
output stream.
compiler/check_promise.m:
compiler/check_typeclass.m:
compiler/closure_analysis.m:
compiler/complexity.m:
compiler/cse_detection.m:
compiler/deforest.m:
compiler/delay_construct.m:
compiler/delay_partial_inst.m:
compiler/deps_map.m:
compiler/direct_arg_in_out.m:
compiler/grab_modules.m:
compiler/handle_options.m:
compiler/hhf.m:
compiler/inlining.m:
compiler/make.module_dep_file.m:
compiler/ml_proc_gen.m:
compiler/ml_top_gen.m:
compiler/mode_constraints.m:
compiler/modes.m:
compiler/polymorphism.m:
compiler/purity.m:
compiler/read_modules.m:
compiler/recompilation.check.m:
compiler/saved_vars.m:
compiler/simplify_proc.m:
compiler/size_prof.m:
compiler/stack_opt.m:
compiler/switch_detection.m:
compiler/typecheck.m:
compiler/unique_modes.m:
compiler/unneeded_code.m:
compiler/write_module_interface_files.m:
Get these modules to take an explicitly specified stream to which
to write progress messages when they are invoked from mercury_compile_*.m.
For predicates in these modules that can be invoked both directly
by mercury_compile_*.m *and* by other modules, the latter effectively
as a subcontractor, make them take a maybe(stream), with the intention
being that all the other modules that use the predicate as a subcontractor
would pass a "no". This avoids the need to pass progress streams
down to the internals of other passes, and also avoids overwhelming
the user invoking the compiler with unnecessary details.
As above, and also delete a progress message that shouldn't be needed
anymore.
Move a test of option value compatibility from
mercury_compile_middle_passes.m to handle_options.m, where it belongs.
compiler/float_regs.m:
Write a debug message to the debug stream.
compiler/pd_info.m:
Include the progress stream in the pd_info structure, because this is
the simplest way to ensure that all parts of the partial deduction pass
have access to it.
compiler/make.build.m:
compiler/make.program_target.m:
compiler/make.track_flags.m:
Make the minimal changes needed to conform to the changes above.
The rest can be done when the make package is converted to consistently
use explicit streams.
compiler/bytecode_gen.m:
compiler/structure_reuse.direct.m:
compiler/structure_reuse.versions.m:
compiler/structure_sharing.analysis.m:
Make the minimal changes needed to conform to the changes above.
The rest can be done when these modules start being maintained again.
compiler/Mercury.options:
Stop specifying --no-warn-implicit-stream-calls for mercury_compile_*.m,
since this diff makes that unnecessary.
Start specifying --no-warn-implicit-stream-calls for some modules that
are not currently being actively maintained, because the addition of
progress-reporting predicates that take explicitly specified streams
would otherwise cause the generation of such warnings for them.
compiler/error_spec.m:
This new module contains the part of the old error_util.m that defines
the error_spec type, and some functions that can help construct pieces
of error_specs. Most modules of the compiler that deal with errors
will need to import only this part of the old error_util.m.
This change also renames the format_component type to format_piece,
which matches our long-standing naming convention for variables containing
(lists of) values of this type.
compiler/write_error_spec.m:
This new module contains the part of the old error_util.m that
writes out error specs, and converts them to strings.
This diff marks as obsolete the versions of predicates that
write out error specs to the current output stream, without
*explicitly* specifying the intended stream.
compiler/error_sort.m:
This new module contains the part of the old error_util.m that
sorts lists of error specs and error msgs.
compiler/error_type_util.m:
This new module contains the part of the old error_util.m that
convert types to format_pieces that generate readable output.
compiler/parse_tree.m:
compiler/notes/compiler_design.html:
Include and document the new modules.
compiler/error_util.m:
The code remaining in the original error_util.m consists of
general utility predicates and functions that don't fit into
any of the modules above.
Delete an unneeded pair of I/O states from the argument list
of a predicate.
compiler/file_util.m:
Move the unable_to_open_file predicate here from error_util.m,
since it belongs here. Mark another predicate that writes
to the current output stream as obsolete.
compiler/hlds_error_util.m:
Mark two predicates that wrote out error_spec to the current output
stream as obsolete, and add versions that take an explicit output stream.
compiler/Mercury.options:
Compile the modules that call the newly obsoleted predicates
with --no-warn-obsolete, for the time being.
compiler/*.m:
Conform to the changes above, mostly by updating import_module
declarations, and renaming format_component to format_piece.
compiler/prog_item.m:
Change the representation of include, import and use declarations
in the data structures of .int0 and .int3 files.
For both .int0 and .int3 files, switch from data structures that can
associate one or more contexts with each such declaration with
data structures that can associate just one. Any duplicates would be
errors, and we don't want to allow the representation of errors in
compiler-generated interface files. (Similar updates to .int/.int2
files are coming soon.)
For .int3 files, delete two fields that hold information about include
declarations and import/use declarations respectively in a fully checked
form. They are not needed, because with invariants that .int3 files
are subject to (no uses, no implementation section, no implicit imports),
they contain just the same info as the new data structures mentioned
in the paragraph above.
compiler/comp_unit_interface.m:
Update the code that computes the contents of .int0/.int3 files.
Fix some out-of-sequence variable names.
compiler/convert_parse_tree.m:
Update the code that converts the generic parsed representation of
interface files to the specific representations of .int0/.int3 files.
compiler/item_util.m:
Update the utility predicates that comp_unit_interface.m and
convert_parse_tree.m use to do their jobs. Add new variants
of some existing predicates.
compiler/grab_modules.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.qualify_items.m:
compiler/parse_tree_out.m:
compiler/recompilation.check.m:
library/term.m:
library/term_context.m:
As above.
Rename the term.context type as term_context.term_context, with
term.context now being defined as an equivalence type.
Replace the context_init function and predicate and the dummy_context_init
function with just one function: dummy_context. This name includes
the important part (the fact that it return a *dummy* context) and deletes
the nonimportant part (dummy contexts are just about never updated,
so the function does not really "initialize" them).
Reduce function/predicate pairs that do the same thing to just a function.
library/MODULES_DOC:
library/library.m:
Add the new module to the list of standard library modules.
NEWS:
Mention the new module, and the obsoleting of the moved predicates
and functions in term.m.
compiler/*.m:
library/*.m:
Conform to the changes above.
The predicates that read a Mercury source, interface or optimization file
used to return four things:
- the name of the file (for callers who specified only a module name),
- the timestamp, if requested and available,
- a parse tree, and
- a representation of any errors (sets of error categories, and error_specs).
However, these four things were not independent. Some combinations of their
values were not allowed, but (a) there was no documentation of what these
combinations were, and (b) code that processed these four things had to be
prepared to handle illegal as well as legal combinations.
This diff makes these predicates return only one result, which contains
- all four of the above things, when the file could be opened, but
- only the file name and a representation of the error if the file
could not be opened,
- only the file name and a representation of *no* errors, if the caller
asked the predicate to read the file only if its timestamp did not match
a specified value, and it does match that value.
We use a somewhat modified version of an existing type, have_read_module,
for this. It is modified both by including information that its users
now need that they did not need before, and shortening the names of its
function symbols, which now occur in many more places than before.
compiler/read_modules.m:
Make the change to the output arguments described above.
Making this change requires having the affected predicates deal with
the case where we either cannot find or cannot open the relevant file.
(The two are effectively the same thing in this module, since we search
by attempting to open.) Passing that task off to parse_modules.m
was always inelegant.
Simplify the have_read_module_map type by making the key always
be a module_name. We deleted the only use of have_read_module_maps
with another kind of key a while ago.
Delete some no-longer-needed predicates.
compiler/parse_module.m:
Delete the code dealing with the absence of a file stream to parse.
Replace code that used to construct dummy parse trees with code
that simply returns *no* parse tree if the file could be opened
but could not be read, or if its timestamp indicated that reading it
would have been redundant.
Delete some utility predicates moved to parse_error.m, so that
read_modules.m could also use them.
Fix comment rot.
compiler/parse_error.m:
Add some utility predicates for use by read_modules.m and parse_module.m.
compiler/deps_map.m:
Conform to the changes above.
Document the dubious effects of an old design decision.
Fix a misleading predicate name.
Fix comment rot.
compiler/grab_modules.m:
Conform to the changes above.
Some predicates used to return parse trees that could be dummies.
Change them to return just the parts of the parse tree that the
caller was interested in, which was usually a tiny part, and which
can be constructed trivially even when we don't have a parse tree.
Delete an unneeded type.
compiler/recompilation.check.m:
Conform to the changes above.
Represent the operations we need to test version numbers in interface files
as a typeclass, and rewrite the checking operation in terms of that
typeclass, with one instance for each kind of interface file.
Move some repeated code into a predicate.
Shorten some long names.
compiler/mercury_compile_main.m:
Conform to the changes above.
Break up a large predicate.
compiler/generate_dep_d_files.m:
compiler/make.module_dep_file.m:
compiler/write_module_interface_files.m:
Conform to the changes above.
compiler/prog_item.m:
Delete the parse_tree_some_int type. The change in recompilation.check.m,
deleted its last few uses there, and allowed the deletion of the last
uses in read_modules.m.
tests/recompilation/two_module_debug:
Extend this script to help deal with problems in all stages of the
execution of a test case, since that was required while debugging
this diff.
Document which parts of this script correspond to which parts of
two_module_test.
tests/recompilation/test_functions:
tests/recompilation/two_module_test:
Simplify the logic of the main test function, by testing a condition
once instead of three times.
Specify whether a recompilation test is expected to succeed or fail
directly, using alternatives that model a bespoke type, instead
of alternatives that mimic a boolean answer to a question, which
does not help readers who don't remember the question.
Always put shell variable names in braces.
Note a problem that makes one of the shell functions ineffective.
tests/recompilation/Mmakefile:
Tell two_module_test what to do using its updated interface.
... and bundle sets of error kinds and their associated lists of error_specs
into a single structure.
compiler/parse_error.m:
Use separate types to represent fatal errors and nonfatal errors,
because those two kinds of errors are often treated differently.
Change the read_module_errors type, which used to be just a set of errors,
to being a set of fatal errors, a set of nonfatal errors, the error
message for each, and error messages for warnings. This preserves
distinctions that we should be able to use in the future to improve
error handling.
Provide utility routines for operating on this now-structured type.
compiler/parse_types.m:
Change the definition of the item_or_marker type slighly to preserve
distinctions that the read_module_errors type now cares about.
compiler/parse_module.m:
Thread the new read_module_errors structure through the predicates
that parse source, interface and optimization files.
Treat the failure of reading the contents of a file as a fatal error.
Previously, we treated it as a nonfatal error.
Delete some fields from types when those field were always set to the
same value, and therefore contained no useful information.
compiler/module_baggage.m:
Switch to using the new read_module_errors structure in module baggage.
compiler/grab_modules.m:
Conform to the changes above.
Use a bespoke type to represent the absence/presence of errors
when reading in optimization files.
compiler/mercury_compile_make_hlds.m:
Conform to the changes above.
Delete a call to pre_hlds_write_out_errors that duplicates
another call a few lines above.
compiler/deps_map.m:
compiler/generate_dep_d_files.m:
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/parse_item.m:
compiler/parse_pragma.m:
compiler/parse_type_defn.m:
compiler/read_modules.m:
compiler/recompilation.check.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
Conform to the changes above.
compiler/module_baggage.m:
This reflects the changes to this module over the last year.
compiler/parse_tree.m:
Change the module name in the include_module declaration.
compiler/deps_map.m:
compiler/generate_dep_d_files.m:
compiler/grab_modules.m:
compiler/make.dependencies.m:
compiler/make.module_dep_file.m:
compiler/make.module_target.m:
compiler/mercury_compile_main.m:
compiler/module_dep_info.m:
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
compiler/recompilation.used_file.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
Change the module name in import_module declarations.
compiler/notes/compiler_design.html:
Change the module name in the module description.
compiler/read_modules.m:
When reading in Mercury source files, interface files or optimization
files, we optionally generate progress messages of the form
% Reading module xyz... done.
But the form of these messages has not been consistent, so that
when reading .opt files, we got messages of the form
% Reading xyz.opt...
% Done.
And we sometimes got messages such as
% Reading .int file for module xyz.int... done.
This is mostly because responsibility for constructing these messages
has been dispersed; different parts of it were generated in different
places, and sometimes when one place switches to a new scheme, not all
the corresponding places where updated accordingly.
This diff changes that by constructing these messages in two predicates
(output_read_msg and output_read_done_msg) in this module, leaving to
callers only a choice between a set of four message forms.
To make this possible, move the read_module_{plain,trans}_opt predicates
here from grab_modules.m. Their equivalents for reading Mercury source
files and interface files were here already, so here is where they belong.
Add several "XXX CLEANUP" markers for future changes.
compiler/parse_module.m:
Do not expect callers to compute the filenames containing .opt
and .trans_opt files, since the code here already does that.
compiler/grab_modules.m:
Delete the predicates moved to read_modules.m (under new names, to fit in
with the predicates already there).
Delete some obsolete comments.
Fix some old variable names.
compiler/deps_map.m:
compiler/generate_dep_d_files.m:
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/recompilation.check.m:
compiler/write_module_interface_files.m:
Conform to the changes above.
compiler/error_util.m:
The internal operation of formatting error messages requires knowing
whether an error_msg should be treated as the first message, or not.
(First messages get indented 1 space; other messages get indented by 3.)
We use the treat_at_first type for this.
In the past, we also used this type in error_msgs themselves, but there,
one of the alternatives of the type, do_not_treat_as_first, is misleading.
Such messages will in fact be treated as first if they are in fact
the first error_msg in an error_spec, which happens quite frequently,
since *most* error_specs contain only one error_msg.
This diff therefore defines a new type, always_treat_as_first,
whose two values have names that *accurately* reflect their meaning:
always_treat_as_first, and treat_based_on_posn. Make all code outside
error_util.m itself use this type; restrict the use of the treat_as_first
type to just the code that does error message formatting.
compiler/add_class.m:
compiler/add_foreign_proc.m:
compiler/check_typeclass.m:
compiler/common.m:
compiler/compile_target_code.m:
compiler/compiler_util.m:
compiler/compute_grade.m:
compiler/fact_table.m:
compiler/find_module.m:
compiler/grab_modules.m:
compiler/handle_options.m:
compiler/make.top_level.m:
compiler/mercury_compile_main.m:
compiler/mercury_compile_mlds_back_end.m:
compiler/mode_errors.m:
compiler/modes.m:
compiler/options_file.m:
compiler/parse_module.m:
compiler/post_typecheck.m:
compiler/prog_event.m:
compiler/recompilation.check.m:
compiler/term_constr_errors.m:
compiler/term_errors.m:
compiler/typecheck_msgs.m:
compiler/write_module_interface_files.m:
Conform to the change in error_util.m. In many cases where this is
appropriate, use the recently introduced simplest_no_context_specs.
NEWS:
Mention this fact.
Group related changes together.
Fix some typos.
compiler/error_util.m:
Let the halt_at_invalid_interface option govern whether we print error
messages when we generate .int/.int2 files, rather than
the print_errors_warnings_when_generating_interface option.
This simplifies the use of that option.
compiler/options.m:
doc/user_guide.texi:
Document the halt_at_invalid_interface option.
Delete the print_errors_warnings_when_generating_interface option,
since its functionality has been subsumed into halt_at_invalid_interface.
compiler/grab_modules.m:
Check the accessibility of imported modules not just when generating target
language code, but also when generating .int/.int2 files. Without this,
the test_nested test case, which this diff moves from invalid to
invalid_make_int, would miss out on some errors being reported
in its new home.
To make the above possible, refactor the code that does the check
to let it work on aug_make_int_units as well as from aug_compilation_units.
Make the wording of any error message depend on where the info came from,
because unlike aug_comp_units, aug_make_int_units work with an
known-incomplete picture of the relevant info.
Update some obsolete comments.
compiler/write_module_interface_files.m:
Add a comment about a design decision that affects this diff.
tests/invalid_make_int/Mercury.options:
tests/invalid_make_int/Mmakefile:
Add to this directory's list the test cases that this diff moves
to this directory from other test directories, because the errors
that they test for are now reported at interface generation time.
Fix the rule handling multimodule tests, now that we have some :-(
Specify -j1 for this directory, since the two tests moved here
from invalid_submodules include nested submodules.
tests/invalid_make_int/bad_type_class_constraint_intermodule.{m,int_err_exp}:
Move this test case here from invalid, renaming files slightly,
and update the expected output.
tests/invalid_make_int/bug499.{m,int_err_exp}:
Move this test case here from invalid, and update the expected output.
tests/invalid_make_int/int_impl_imports.{m,int_err_exp}:
tests/invalid_make_int/int_impl_imports_2.m:
Move this test case here from invalid, and update the expected output.
tests/invalid_make_int/missing_interface_import2.{m,int_err_exp}:
tests/invalid_make_int/missing_interface_import3.m:
Move this test case here from invalid, and update the expected output.
tests/invalid_make_int/missing_parent_import.{m,int_err_exp}:
tests/invalid_make_int/children.m:
tests/invalid_make_int/children2.m:
Move this test case here from invalid_submodules,
and update the expected output.
tests/invalid_make_int/sub_c.{m,int_err_exp}:
tests/invalid_make_int/sub_a.m:
Move this test case here from invalid_submodules,
and update the expected output.
tests/invalid_make_int/test_nested.{m,int_err_exp}:
tests/invalid_make_int/parent.m:
tests/invalid_make_int/parent.private_child.m:
tests/invalid_make_int/parent.public_child.m:
tests/invalid_make_int/parent.undeclared_child.m:
tests/invalid_make_int/parent2.child.m:
tests/invalid_make_int/parent2.m:
Move this test case here from invalid, and update the expected output.
tests/invalid_make_int/transitive_import.{m,int_err_exp}:
Move this test case here from invalid, update it,
and update the expected output.
tests/invalid/Mmakefile:
Delete from the list of tests in this directory the tests
moved to invalid_make_int, and test bug521.
tests/invalid/bug521.{m,err_exp}:
tests/invalid/bug521_sub.m:
Delete this test. invalid_make_int already had a copy of the relevant
part of this test.
tests/invalid_submodules/Mmakefile:
Delete from the list of tests in this directory the two tests
moved to invalid_make_int.
Delete from the list of tests in this directory two other tests
for which we now detect some or all of the errors tested for
at interface generation time, but which are already covered
by other tests.
tests/valid_make_int/Mmakefile:
Delete from the list of tests in this directory the test
moved to invalid_make_int. (We could successfully generate
a .int file for that test case *only if* we did not check it
for errors :-)
tests/valid_seq/intermod_nested_module_bug2.m:
tests/valid_seq/intermod_nested_module_bug2.sub.m:
Move two module imports from the parent to the child module,
because only the child needs them, and any compiler with this diff
will now complain about them being unused in the parent
at interface generation time. Delete a third import, which was
not used anywhere.
This means that we can do the exact same checks (and, if needed, generate
the exact same error messages) when generating interface files as when
generating target language code.
This should also allow us to simplify the process of adding type, inst
and mode definitions to the HLDS.
compiler/prog_item.m:
As above.
Delete unused function.
compiler/error_util.m:
Add mechanisms that allow us to distinguish (a) error specs that represent
a type, inst or mode name definition being invalid, from (b) other error
specs.
compiler/check_type_inst_mode_defns.m:
Improve the error messages we generate, in several ways.
First, for each message, specify a real severity. When the messages
could be seen only when generating interface files, making them all
warnings was fine and preserved old behavior, but since soon these
will be the only place for these checks, we need to call errors errors.
Second, specify, in the phase, which errors represent a invalid type,
inst or mode definition, and which don't.
Third, improve the wording of messages. In some places, do this by
being clearer about the distinction between declarations and definitions.
In others, do it by including more information in the message. In yet
others, do it by recovering some kinds of mistakes (mostly items being
in the wrong section) enough to avoid avalanche errors.
Fourth, fix a bug. If a type ctor has an exported *declaration*,
then it is ok for any foreign type definitions for that type_ctor
being in the implementation section, but if the type_ctor has an
exported Mercury *definition*, then any foreign_type definitions
must be in the interface section as well. The code that handled
both these situations did not enforce that.
Fifth, fix another bug: do not include foreign type definitions
in the source definitions of a type_ctor twice, once as a "du" (sic)
definition, and once as itself.
compiler/convert_parse_tree.m:
Check type, inst and mode definitions in raw_compilation_units
when creating parse_tree_module_srcs.
compiler/comp_unit_interface.m:
Make the code constructing interface files work from the checked maps
in parse_tree_module_srcs.
compiler/make_hlds_passes.m:
Set the flags that say we have found invalid type, inst or mode definitions
from the error specs constructed during the creation of the checked
type, inst and mode maps.
compiler/add_type.m:
Comment out an error message for a condition that will be detected and
reported by check_type_inst_mode_defns.m.
compiler/make_hlds_separate_items.m:
For now, turn checked maps of type, inst and mode definitions
back to item lists for addition to the HLDS. Adding whole checked
definitions to the HLDS will be done by a future change.
compiler/make_hlds_error.m:
Fix misleading indentation in an error message.
compiler/parse_tree_out.m:
Conform to the changes above.
Generate output whose indentation does not depend on tab settings.
compiler/check_raw_comp_unit.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.qualify_items.m:
Conform to the changes above.
compiler/parse_type_defn.m:
Fix misleading error message for ":- type <name> = <type>".
tests/debugger/foreign_type.{m,exp}:
Delete a redundant type declaration to avoid a warning, and update
the .exp file to expect the new line numbers.
tests/invalid/any_mode.err_exp:
tests/invalid/bug436.err_exp:
tests/invalid/bug476.err_exp:
tests/invalid/exported_foreign_enum.err_exp:
tests/invalid/fe_unmapped_nonverbose.err_exp:
tests/invalid/fe_unmapped_verbose.err_exp:
tests/invalid/foreign_enum_invalid.err_exp:
tests/invalid/foreign_solver_type.err_exp:
tests/invalid/foreign_type_visibility.err_exp:
tests/invalid/pragma_qual_error.err_exp:
tests/invalid/repeated_field_name.err_exp:
tests/invalid/subtype_foreign.err_exp:
tests/invalid/type_with_no_defn.err_exp:
tests/invalid/types2.err_exp:
tests/invalid/user_field_access_decl_conflict.err_exp:
tests/invalid_nodepend/bad_foreign_type.err_exp:
tests/invalid_nodepend/bigtest.err_exp:
tests/invalid_nodepend/invalid_typeclass.err_exp:
tests/invalid_nodepend/types.err_exp:
tests/invalid_nodepend/uu_type.err_exp:
tests/invalid_nodepend/where_abstract_enum.err_exp:
Expect the new error messages.
tests/invalid/abstract_solver_type.{m,err_exp}:
tests/warnings/abstract_solver_type.{m,exp}:
Move the abstract_solver_type test case from invalid to warnings, because
this diff changes its only error to be only a warning.
tests/invalid/Mmakefile
compiler/prog_item.m:
As above.
compiler/comp_unit_interface.m:
compiler/convert_parse_tree.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/parse_tree_out.m:
compiler/recompilation.check.m:
compiler/recompilation.version.m:
Conform to the change above.
With this change, we guarantee that if any type, inst or mode
has inconsistent definitions in its defining module, those inconsistencies
will not make it into the the module's .int/.int2/.int3 files, where
they would confuse any compiler invocation that reads them in.
compiler/prog_item.m:
Replace {type,inst,mode}_ctor_defn_maps, which are unchecked,
with {type,inst,mode}_ctor_checked_maps in parse_tree_int[123]s.
To make this possible,
- generalize the items containing inst and mode definitions the way that
items containing type definitions have been generalized previously,
to allow them to store particular *kinds* of inst or mode definitions;
- move the definitions of type_ctor_checked_map and its components
here from check_type_inst_mode_defns.m; and
- add similar, but simpler, definitions for {inst,mode}_ctor_checked_map.
compiler/check_type_inst_mode_defns.m:
Delete the type definitions moved to prog_item.m.
Modify the checking process slightly to allow it to check properly
the definitions we now put into .int, .int2 and .int3 files.
Add code to check inst and mode definitions as well as type definitions.
Since insts and modes have only one non-abstract kind of definition,
these codes are much simpler than the code for checking types.
compiler/comp_unit_interface.m:
Construct checked type, inst and mode definitions to put into
.int, .int2 and .int3 files. The .int2 and .int3 parts were
reasonably simple, but the .int part requires some detailed
case-by-case analysis.
compiler/convert_parse_tree.m:
Check the type, inst and mode definitions read in from .int/.int2/.int3
files when creating their parse trees. If those files were generated
by a compiler that has this diff, then this checking process should
not find any problems.
compiler/equiv_type.m:
Operate on checked type, inst and mode definitions instead of their
unchecked versions.
Delete an unneeded field from a structure.
compiler/error_util.m:
Add {qual,unqual}_{inst,mode}_ctor as format components, so that
any error messages about inst and mode definitions do not have to convert
inst_ctors and mode_ctors to symname/arity pairs.
compiler/item_util.m:
Add some utility predicates and functions needed by the changes above.
compiler/module_qual.collect_mq_info.m:
Since checked type, inst and mode definitions can contain info that is
derived from both the interface and implementation sections of a module,
add the capability to get only the *publicly* declared types, insts and
modes from their respective kinds of checked definitions.
compiler/module_qual.qualify_items.m:
Module qualify checked type, inst and mode definitions.
compiler/module_qual.qual_errors.m:
Treat references to inst_ctors and mode_ctors in qualification error
messages the same way as we treat type_ctors.
Also replace mq_ids that are intended to represent classes with class_ids,
to represent the intent better.
compiler/parse_tree_out.m:
Output the updated parse_trees of .int/.int2/.int3 files.
compiler/prog_data.m:
Add an XXX about a future change.
compiler/decide_type_repn.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/pred_table.m:
compiler/recompilation.check.m:
compiler/recompilation.version.m:
Conform to the changes above.
library/maybe.m:
Add utility predicates needed by the code above.
NEWS:
Announce the new predicates in maybe.m.
tools/intdiffall:
A script that shows differences between the automatically generated
interface files between stage 1 and stage 2, which helped debug this diff.
Also, turn some semidet functions into predicates, and given them
more meaningful names.
compiler/recompilation.m:
compiler/timestamp.m:
As above.
compiler/comp_unit_interface.m:
compiler/grab_modules.m:
compiler/make_hlds_passes.m:
compiler/parse_item.m:
compiler/parse_module.m:
compiler/parse_tree_out.m:
compiler/parse_types.m:
compiler/prog_item.m:
compiler/recompilation.check.m:
compiler/recompilation.m:
compiler/recompilation.usage.m:
compiler/recompilation.version.m:
Conform to the changes above.
compiler/prog_item.m:
Split aug_compilation_unit into two types. One, aug_make_int_unit,
is a new type, and it is a version of the old aug_compilation_unit
that contains only what is needed when making .int{,0,2} files.
The other, the new aug_compilation_unit, contains what is needed
when generating target language code. Both encode in their argument types
stronger invariants that the old aug_compilation_unit type did.
compiler/comp_unit_interface.m:
compiler/decide_type_repn.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.m:
compiler/parse_tree_out.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
Conform to the change above.
In some cases, this means passing around aug_make_int_units instead of
aug_compilation_units.
In a few cases, it means creating two copies of a predicate,
one for aug_compilation_units and one for aug_make_int_units.
In such cases, factor out any commonalities between the copies.
In many cases, it means updates to account for the specialized types
of the fields of aug_compilation_unit or aug_make_int_unit.
In some cases, it means deleting code that was always unreachable,
but the not-tight-enough types did not make this fact visible.
(Due to this effect, this diff removes more lines than it adds.)
compiler/prog_item.m:
Split the parse_tree_module_src that holds the languages for which
the module needs a self-foreign-import into two, depending on the
section (interface or implementation) into which the self-FIM should go.
compiler/convert_parse_tree.m:
When computing the self-FIM languages, put each one into its section.
compiler/comp_unit_interface.m:
Don't compute the interface self-FIMs, since they are now available
in the parse_tree_module_src.
Document why we *do* compute the implementation self-FIMs
for the interface file we are constructing.
compiler/check_raw_comp_unit.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.qualify_items.m:
compiler/parse_tree_out.m:
Conform to the changes above.
compiler/module_imports.m:
Delete the module_and_imports type, and replace its uses with
aug_compilation_units.
The structure of module_and_imports used to be private, but the
structure of aug_compilation_units is public. Delete the getter/setter
predicates on module_and_imports structures that can now be done
using field names on aug_compilation_units. Delete, along with this,
the infrastructure for keeping track of which module_and_imports fields
are ever accessed. It has served its purpose, but it won't be needed
anymore.
Change types that used to include module_and_imports to include
aug_compilation_units instead.
Rename any types, predicates and functions that used to contain
the "module_and_imports" string in their name.
compiler/prog_item.m:
Move the version number map to be the last field of aug_compilation_units,
as it used to be in module_and_imports structures, since this fits better.
compiler/get_dependencies.m:
Make the name of a predicate more reflective of the type it returns.
compiler/comp_unit_interface.m:
compiler/deps_map.m:
compiler/equiv_type.m:
compiler/generate_dep_d_files.m:
compiler/grab_modules.m:
compiler/make.m:
compiler/make.module_dep_file.m:
compiler/make_hlds_separate_items.m:
compiler/mercury_compile_main.m:
compiler/module_qual.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
Conform to the changes above.
compiler/convert_parse_tree.m:
Always compute the set of foreign languages for which
the current module needs a self-import FIM. Use the algorithm
that grab_unqual_imported_modules_make_int used to use to fill in
that slot in the parse_tree_module_src.
Stop computing some values whose only user was a sanity check.
If it hasn't triggered by now, it won't.
Delete an unused predicate that won't be needed.
compiler/prog_item.m:
Since that field in the parse_tree_module_src is now always filled in
when the parse_tree_module_src is created, delete the maybe wrapper
around it.
compiler/grab_modules.m:
Don't fill in the already-filled-in field.
compiler/comp_unit_interface.m:
Conform to the changes above, and add an XXX.
compiler/parse_tree_out.m:
Conform to the changes above.
compiler/module_imports.m:
Keep in the module_and_imports structure only the fields that used to
construct aug_compilation_units, to prepare for the replacement
of module_and_imports by aug_compilation_unit. Move the remaining fields
into a new structure, the module_baggage, and introduce a new type,
module_imports_and_baggage, that contains module_and_imports (for now)
and module_baggage.
The point of this is to allow the splitting of aug_compilation_unit
into types:
- one (aug_compilation_unit itself) for use when generating
target language code, and
- one (maybe named aug_make_int_unit) for use when generating
.int/.int2 files.
These two use cases fill some of the fields, such as direct and indirect
imports, with different kinds of parse trees. Until now, we handled this
using types such as direct_int_spec, which could contain either kind
of parse tree, but having separate aug_compilation_unit/aug_make_int_unit
types would allow the compiler to enforce the different invariants
of the two use cases.
This split requires factoring out the common parts of module_and_imports
structure. The module_baggage type contains these common parts.
Simplify some operations as permitted by the above change.
Rename others to reflect the new names of the types they operate on.
Don't return an output argument that is contained in another output
argument.
Delete operations on deleted fields.
Delete a predicate that has become identical to another.
compiler/grab_modules.m:
Add operations on module_baggage (that previously were operations on
module_and_imports) that are needed only in this module.
Conform to the changes above.
compiler/deps_map.m:
compiler/generate_dep_d_files.m:
compiler/make.m:
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
Conform to the changes above.
compiler/module_imports.m:
Originally, we had three different ways to initialize module_and_imports
structures. In the last few weeks, I replaced one of them with code
that creates a module_dep_summary instead, which left two:
init_module_and_imports and make_module_and_imports. These differed
in how (and, more to the point, whether) they initialized the fields
of the module_and_imports structure that semantically were properties
that were computable from the other fields, the fields that would
eventually be put into the augmented compilation unit. Changes in the
last few weeks have deleted all these "computable from elsewhere" fields,
leaving only a trivial difference between init_module_and_imports
and make_module_and_imports.
This diff makes the last step, and replaces both those predicates
with a single predicate, which is also called init_module_and_imports,
whose argument list is the union of the argument lists of its two
progenitors.
compiler/grab_modules.m:
Conform to the change above.
compiler/module_imports.m:
As above.
compiler/grab_modules.m:
Add indirectly imported module names only to indirect_int_specs field,
and not redundantly to the now-deleted field.
compiler/write_deps_file.m:
Get the same info from the indirect_int_specs field.
compiler/module_imports.m:
Delete module_and_imports' {int,imp}_deps fields, as their contents
are computable from the field containing the parse_tree_module_src.
Add a utility predicate that returns the int and imp dependencies
separately. Clients for whom the int/imp distinction is not relevant
can get the union of those sets more simply by just taking the
key set of the parse_tree_module_src's import_and_or_use_map.
Delete now unneeded arguments.
Make some minor rearrangements to minimize the remaining differences
between the two methods of initializing module_and_imports structures,
init_module_and_imports and make_module_and_imports.
compiler/grab_modules.m:
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/write_deps_file.m:
Conform to the change above.
compiler/module_imports.m:
As above. The new approach has a much simpler correctness argument,
especially since some of the methods for initializing module_and_imports
structures didn't actually fill in one of the fields with meaningful
information.
Don't ask callers for arguments that are not used.
compiler/get_dependencies.m:
Add utility predicates for computing the information that those fields
used to contain on demand.
Simplify the interface to some similar recently-added utility predicates.
compiler/make.module_dep_file.m:
Call those utility predicates when needed.
compiler/item_util.m:
Delete the utility predicates that module_imports.m *used* to use
to fill in those fields, which is no longer needed.
compiler/grab_modules.m:
compiler/write_deps_file.m:
Conform to the changes above.
compiler/module_imports.m:
Delete the field containing foreign_include_file_infos from
the module_and_imports structure, because it is redundant;
the same info is implicit in the parse_tree_module_src field.
compiler/grab_modules.m:
Don't pass the set of foreign_file_include_infos to module_imports.m,
since it does not need it anymore.
compiler/get_dependencies.m:
Provide a predicate to get the set of foreign_file_include_infos
on demand. Provide this info in the form of a set, not a cord,
since duplicates don't count.
compiler/make.module_dep_file.m:
compiler/write_deps_file.m:
Get the set of foreign_file_include_infos when needed.
compiler/prog_data_foreign.m:
Delete a type synonym that is not needed anymore.
compiler/module_imports.m:
Delete the redundant fact table field from module_and_imports.
Get that info when needed from the parse_tree_module_src stored
in the module_and_imports structure.
compiler/get_dependencies.m:
Add a predicate to get this info on demand.
compiler/grab_modules.m:
Don't precompute the set of fact tables. (Actually, for now we do
precompute it but ignore it, but the precomputation should be gone soon.)
compiler/make.module_dep_file.m:
Conform to the changes above, and give a predicate a better name.
compiler/write_deps_file.m:
Conform to the changes above.
compiler/grab_modules.m
The recent failures in tests/invalid on testing.mercurylang.org
occurred with intermodule optimization being enabled, and were due to
not being able to find the .int file of the module being compiled.
The bug occurred only with intermodule optimization because the call
to read in the problematic .int file was in the predicate that grabbed
.opt files, .trans_opt files and the .int* files they needed, NOT
in its proper place, which is in the predicate that grabs all the files
needed to create an augmented compilation unit to be compiled.
Fix this my moving the call.
Bootchecks with --intermod-opt in Mmake.stage.params were not affected
because by default, tools/bootcheck does NOT execute the test cases
with the flags in Mmake.stage.params. We should maybe change that
default.
With the moved call, all compilations will try to read in the module's
own .int file. This is what we ordinarily want. However, it is not what
we want in tests/invalid, because
- in multi-module programs, for which we make dependencies, and for which
we should build those .int files, it can change the symptoms of errors,
and
- in single-module programs, for which we do not (now) make dependencies,
we totally replace all error with the error that the .int file cannot
be read. For many single-module programs, making dependencies would
not be a cure either, because the bugs we deliberately put into their
source code would cause the process of making dependencies itself fail.
Likewise, the tests in tests/recompilation fail with this change,
because the test scripts have not yet been updated to generate the
.int file.
For these reasons, the code to read in the module's own .int file
is now guarded by an option, experiment3, which defaults to off.
I intend that this guard be removed once the problems listed above
have been fixed. The code inside the guard also describes some
required future work.
The movement of the call exposed two other bugs, which this diff
also fixes.
The first bug is that for modules that have submodules, such as
string.m in the library dir, it was the presence of the call
in its old position that prevented the following code from
trying to read in the module's own .int0 file. Reading in that file
leads to a duplicate definition error for everything defined in the
module, with one definition in the .m file, and another in the .int0
file derived from it. This is clearly bad, but the fix is simple:
exclude the module being compiled from having its .int0 file
being read in. (Its ancestors were already excluded.)
The second bug is that the kinds of errors that we do not report
while *constructing* the .int file (which is all of them for now,
since we suppress error message generation at that time by default)
but which we do report when *reading* the .int file, the inclusion
of those error reports causes mismatches against the expected error
output files in tests/invalid. This diff thus does not add the error specs
from reading in .int files to the module_and_imports structure
(which effectively means forgetting about them altogether) if the reason
for reading the .int file was getting type representation information.
compiler/prog_item.m:
Delete the module name and module name context fields from
aug_compilation_units, since the same info is available from
the parse_tree_module_src field.
compiler/comp_unit_interface.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/grab_modules.m:
compiler/item_util.m:
compiler/make_hlds_passes.m:
compiler/make_hlds_separate_items.m:
compiler/mercury_compile_main.m:
compiler/module_imports.m:
compiler/module_qual.m:
compiler/prog_item_stats.m:
compiler/write_deps_file.m:
Conform to the change above.
compiler/module_imports.m:
As above; delete the ancestors field, since users can compute its value
by calling get_ancestors_set on the module name.
On some initialization paths, the ancestors field used to be filled
with a dummy empty set, so this change should improve robustness.
compiler/deps_map.m:
compiler/grab_modules.m:
compiler/make.dependencies.m:
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/module_deps_graph.m:
compiler/write_deps_file.m:
Conform to the changes in module_imports.m.
... not from separate fields in the module_and_imports_structure.
compiler/module_imports.m:
As above. This is a step towards using the parse_tree_module_src
as the single source of truth about the properties of the module
being compiled.
compiler/deps_map.m:
compiler/grab_modules.m:
compiler/write_deps_file.m:
Conform to the changes in module_imports.m.
compiler/module_imports.m:
The module_and_imports structure used to have a field, mf_nested_children,
which stored a set of module names, but which had confusing semantics:
it had two different meanings in two different circumstances.
Replace that field with a value of a new type, maybe_top_module, which
has two separate function symbols for those two circumstances:
when a module is the top module in its source file, and when it is not.
compiler/mercury_compile_main.m:
Conform to the change above.
Put related arguments next to each other.
compiler/grab_modules.m:
compiler/make.dependencies.m:
compiler/make.module_dep_file.m:
compiler/make.module_target.m:
compiler/recompilation.check.m:
compiler/write_deps_file.m:
Conform to the change above.
tests/recompilation/test_functions:
Impose a limit of 10 minutes of CPU time on each process. This should
help catch infinite loops, such as happened during the development
of this diff.
Fix indentation, and clarify some comments.
tests/recompilation/README:
Fix stray blank line.
... instead of into raw_compilation_units. Besides being more convenient to
work on, a parse_tree_module_src encodes in its type a significant number
of invariants that a raw_compilation unit does not.
compiler/split_parse_tree_src.m:
Immediately after creating a raw_compilation_unit for one of the modules
in a potentially multi-module source file, convert it to a
parse_tree_module_src.
Fix an old problem that can come up in contrived erroneous code.
Specifically, when a submodule is both nested inside its parent module,
*and* it also has an explicit include_module declaration, we used to
record the contexts of both "inclusions", after generating an error
message. The more rigorous checking that later code now does on the
resulting inclusion map would look at this "duplicate inclusion"
and generate *another* error message. To avoid this redundant message,
do not record the contexts of erroneous inclusions for which a message
has already been generated.
compiler/grab_modules.m:
Operate on parse_tree_module_srcs instead of raw_compilation_units.
This allows us to avoid having code for doing what converting the
raw_compilation_unit to a parse_tree_module_src has already done.
In fact, that conversion code does a better job. The old code assumed
that all implicitly available modules are used in the interface,
whereas in fact only some should be used in the interface, with
the rest being used in the implementation section.
compiler/module_imports.m:
Make the predicates that create module_and_imports structures
take a parse_tree_module_src instead of a raw_compilation_unit
as input. For now, we convert the parse_tree_module_src back to
a raw_compilation_unit for further processing, but I intend
a later diff to change this. Nevertheless, one immediate change
is that init_module_and_imports now stores the *actual*
parse_tree_module_src in the module_and_imports structure,
not a dummy.
compiler/prog_item.m:
Do not include a list of foreign_enum items in the interface section
of a parse_tree_module_src, since such items are not allowed to occur
in interface sections.
For the same reason, delete the field for foreign_enums in the interface
sections of .int0 and .int files.
compiler/check_raw_comp_unit.m:
Operate on parse_tree_module_srcs instead of raw_compilation_units.
compiler/comp_unit_interface.m:
Operate on parse_tree_module_srcs instead of raw_compilation_units.
Do not expect any foreign_enum items in interface sections, since
they are not allowed there.
compiler/read_modules.m:
Provide a mechanism to remember having read a parse_tree_module_src.
compiler/write_module_interface_files.m:
Operate on parse_tree_module_srcs instead of raw_compilation_units.
compiler/get_dependencies.m:
Add a new version of an existing utility predicate. The old one operated
on raw item lists, the new one operates on parse_tree_module_srcs.
To make this possible, factor out the code pieces that operate on
each non-ignored kind of item.
compiler/item_util.m:
Add some new utility predicates/functions.
compiler/convert_parse_tree.m:
Conform to the change in prog_item.m. (We already generated an error
message for a foreign_enum item in the interface, but still passed around
a list of foreign_enum items that was guaranteed to be stay empty.)
compiler/make.module_dep_file.m:
Conform to the changes above.
Use explicit streams in one place.
Do not pass an unneeded argument.
compiler/check_parse_tree_type_defns.m:
compiler/equiv_type.m:
compiler/make_hlds_separate_items.m:
Conform to the change in prog_item.m.
compiler/mercury_compile_main.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.qualify_items.m:
compiler/parse_tree_out.m:
Conform to the changes above.
library/map.m:
library/tree234.m:
Add foldl6, foldl6_values and foldr6 predicates. An earlier version
of this diff needed foldl6, and I added the others for symmetry.
NEWS:
Announce the new library predicates.
tests/invalid/bad_mutable.m:
Export something, to avoid a warning about not exporting anything.
tests/invalid_submodules/duplicate_module.m:
tests/invalid_submodules/duplicate_module_test.m:
Update programming style.
compiler/du_type_layout.m:
Add a first draft of the code that will eventually get type representation
information from .int files. Invoke the new code only if undocumented,
developer-only experiment flags are set, and when invoked, compare
its output with the output of the old algorithm it is intended to
eventually replace.
The new code is not yet complete, and thus has not yet been tested.
compiler/prog_item.m:
Add a new field to the augmented compilation unit data type
that is intended to hold the .int file of the main module, the module
that the augmented compilation unit is for. For every module imported
by the main module, we read its .int or .int2 file and get the
representation information for the types they define from those files,
but we need this info for the main module's own types as well.
Add "need type representation information" as reason for reading
a .int file.
Break some types down into smaller types, to allow their components
to be manipulated individually by du_type_layout.m.
Start using uint8, not uint, in situations where we want to store
values that select bits in words, which fit into 6 bits.
Add XXXs where we could, and maybe should switch to uint8s.
Add extra documentation.
compiler/grab_modules.m:
Read in the main module's own .int file.
Replace a bool with a bespoke type.
compiler/module_imports.m:
Provide a slot for the main module's own interface file,
which will be used to set the new slot in augmented compilation units.
compiler/comp_unit_interface.m:
Put the exact same type_ctor_repn_map into .int2 files as .int files.
This should prevent the scenario where a compilation reads
e.g. mod_x.int2, gains access to the definitions of the exported types
of module mod_x, but then needs to read mod_x.int as well to find out
how those types are represented. Since neither type checking nor any
other semantic analysis pass pays any attention to type_repn items,
the additional type_repn items in mod_x.int2 cannot turn an otherwise
illegal program into a legal one.
compiler/hlds_data.m:
Clarify a comment.
compiler/maybe_error.m:
Provide a utility predicate.
compiler/prog_data.m:
Add XXXs about possible future improvements.
compiler/decide_type_repn.m:
compiler/equiv_type.m:
compiler/get_dependencies.m:
compiler/make_hlds_passes.m:
compiler/make_hlds_separate_items.m:
compiler/module_qual.collect_mq_info.m:
compiler/module_qual.m:
compiler/parse_tree_out.m:
compiler/parse_tree_out_type_repn.m:
compiler/parse_type_repn.m:
compiler/prog_item_stats.m:
compiler/write_deps_file.m:
Conform to the changes above.
That info hasn't been used in a long time (since we deleted the IL backend,
according to an old comment).
compiler/module_imports.m:
Delete the has_main field from the module_and_imports structure.
compiler/convert_parse_tree.m:
compiler/get_dependencies.m:
Don't look for main/2.
compiler/make.module_dep_file.m:
When reading in module_dep structures, ignore the contents of the
no_main/has_main field. The value in that field used to be stored
elsewhere, but it was never used.
When writing out module_dep structures, always put "no_main"
into the no_main/has_main field.
Add comments describing possible changes to the on-disk module_dep
structure.
compiler/grab_modules.m:
Conform to the changes above.
This implements Mantis feature request #495.
NEWS:
Announce the change.
compiler/optimization_options.m:
A new module for managing optimization options.
It defines a separate bespoke type for every boolean optimization option
to make it harder to confuse them. It defines a tuple type (opt_tuple)
for accessing optimization options quickly. It implements the turning on
(but NOT turning off) of optimizations when a given optimization level
is selected.
tools/make_optimization_options_middle:
tools/make_optimization_options_db:
The script that generates the meat of optimization_options.m,
and the database of option names, kinds and initial values
that it uses as its input. The script also generates some code
for the special_handler predicate in compiler/options.m.
tools/make_optimization_options_start:
tools/make_optimization_options_end:
The handwritten initial and final parts of optimization_options.m.
tools/make_optimization_options:
The script that pulls these parts together to form optimization_options.m.
compiler/options.m:
Make every optimization option a special option, to be handled by
the special_handler predicate. That handling consists of simply
adding a representation of the option to the end of a cord of
optimization options, to be processed later by optimization_options.m.
That processing will record the values of these options in the opt_tuple,
which is where every other part of the compiler should get them from.
Change the interface of special_handler to make the above possible.
Add an "optopt_" (optimization option) prefix to the name of
every optimization option, to make them inaccessible to the rest
of the compiler under their old name, and thus help enforce the switch
to using the opt_tuple. Any access to these options to look up
their values would fail anyway, since the option data would no longer be
e.g. bool(yes), but bool_special, but the name change makes this failure
happen at compile time, not runtime.
Reclassify a few options to make the above make sense. Some options
(unneeded_code_debug, unneeded_code_debug_pred_name, and
common_struct_preds) were classified as oc_opt even though they
control only the *debugging* of optimizations, while some options
(c_optimize and inline_alloc) were not classified as oc_opt
even though we do set them automatically at some optimization levels.
Delete the opt_level_number option, since it was not used anywhere.
Delete the code for handling -ON and --opt-space, since that is now
done in optimization_options.m.
Add some XXXs.
compiler/handle_options.m:
Switch to using getopt_io.process_options_userdata_se, as required
by the new interface of the special_handler in options.m.
In the absence of errors, invoke optimization_options.m to initialize
the opt_tuple. Then update the opt_tuple incrementally when processing
option implications that affect optimization options.
compiler/globals.m:
Put the opt_tuple into a new field of the globals structure.
compiler/accumulator.m:
compiler/add_pragma_type_spec.m:
compiler/add_trail_ops.m:
compiler/code_info.m:
compiler/code_loc_dep.m:
compiler/compile_target_code.m:
compiler/const_struct.m:
compiler/deforest.m:
compiler/dep_par_conj.m:
compiler/disj_gen.m:
compiler/erl_code_gen.m:
compiler/format_call.m:
compiler/global_data.m:
compiler/grab_modules.m:
compiler/higher_order.m:
compiler/hlds_pred.m:
compiler/inlining.m:
compiler/intermod.m:
compiler/ite_gen.m:
compiler/jumpopt.m:
compiler/libs.m:
compiler/llds_out_code_addr.m:
compiler/llds_out_data.m:
compiler/llds_out_file.m:
compiler/llds_out_instr.m:
compiler/llds_out_util.m:
compiler/matching.m:
compiler/mercury_compile_front_end.m:
compiler/mercury_compile_llds_back_end.m:
compiler/mercury_compile_main.m:
compiler/mercury_compile_middle_passes.m:
compiler/mercury_compile_mlds_back_end.m:
compiler/ml_disj_gen.m:
compiler/ml_gen_info.m:
compiler/ml_lookup_switch.m:
compiler/ml_optimize.m:
compiler/ml_proc_gen.m:
compiler/ml_simplify_switch.m:
compiler/ml_switch_gen.m:
compiler/ml_unify_gen_construct.m:
compiler/optimize.m:
compiler/pd_util.m:
compiler/peephole.m:
compiler/polymorphism.m:
compiler/proc_gen.m:
compiler/simplify_goal_call.m:
compiler/simplify_goal_scope.m:
compiler/simplify_info.m:
compiler/simplify_proc.m:
compiler/simplify_tasks.m:
compiler/stack_layout.m:
compiler/stack_opt.m:
compiler/switch_gen.m:
compiler/switch_util.m:
compiler/tag_switch.m:
compiler/tupling.m:
compiler/unify_gen_construct.m:
compiler/unneeded_code.m:
compiler/unused_args.m:
Conform to the changes above, mostly by looking up optimization options
in the opt_tuple. In some places, replace bools containing optimization
options with the bespoke type of that specific optimization option.
library/getopt_template:
Fix a bug that screwed up an error message.
The bug happened when processing a --file option. If one of the
options in the file was a special option whose special handler failed,
the code handling that failing option returned both an error indication,
and the rest of the argument list read in from the file. The code
handling the --file option then *ignored* the error indication from
the failed special option, and returned an error message of its own
complaining about the unconsumed remaining arguments in the file,
believing them to be non-option arguments, even though these arguments
were never looked it to see if they were options.
The fix is for the code handling --flag options to check whether
the code processing the file contents found any errors, and if so,
return that error *without* looking at the list of remaining arguments.
In an unrelated change, factor out a duplicate call.
This is a first step towards a much finer grained partition.
compiler/file_names.m:
Split the ext type into ext_src and ext_other, as mentioned above.
Add the first predicate for checking whether a string falls into
a given category of extensions.
Add an XXX proposing a better solution for an old problem that does not
actually arise in practice.
compiler/compile_target_code.m:
Split the two-moded predicate maybe_pic_object_file_extension into
two separate one-mode predicates, one for each old mode. The
implementations of the two modes were already separate, because
the two modes already did different jobs: while one went from PIC
to an "extension", the other went from an "extension string" to PIC.
Until now, "extension" and "extension string" were equivalent;
after this diff, they aren't anymore.
Delete an unused argument.
compiler/make.util.m:
Split the two-moded predicate target_extension into
two separate one-mode predicates, one for each old mode,
for the same reason as maybe_pic_object_file_extension above:
the fact that "extension" and "extension string" are now distinct.
compiler/options_file.m:
Move debug infrastructure here from mercury_compile_main.m, to help
debug possible problems with options files. (I had such a problem
while writing this diff.)
Improve how progress messages are printed.
compiler/options.m:
Make an error message more useful.
compiler/mercury_compile_main.m:
Add infrastructure for debugging possible problems with command lines.
(I had such a problem while writing this diff.)
compiler/analysis.m:
Conform to the changes above. Put the arguments of some methods
into the same order as similar predicates in file_names.m.
compiler/find_module.m:
Conform to the changes above. Delete an unused argument,
compiler/analysis.file.m:
compiler/du_type_layout.m:
compiler/elds_to_erlang.m:
compiler/export.m:
compiler/fact_table.m:
compiler/file_kind.m:
compiler/generate_dep_d_files.m:
compiler/grab_modules.m:
compiler/llds_out_file.m:
compiler/make.build.m:
compiler/make.deps_set.m:
compiler/make.m:
compiler/make.module_dep_file.m:
compiler/make.module_target.m:
compiler/make.program_target.m:
compiler/mercury_compile_front_end.m:
compiler/mercury_compile_llds_back_end.m:
compiler/mercury_compile_middle_passes.m:
compiler/mercury_compile_mlds_back_end.m:
compiler/mlds_to_c_file.m:
compiler/mlds_to_cs_file.m:
compiler/mlds_to_java_file.m:
compiler/mmc_analysis.m:
compiler/mode_constraints.m:
compiler/module_cmds.m:
compiler/prog_foreign.m:
compiler/read_modules.m:
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
compiler/source_file_map.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
compiler/xml_documentation.m:
... to allow later changes to its definition.
compiler/file_names.m:
We used to represent filename extensions simply as strings. This meant
all calls to the predicates in file_names.m that convert module names
to file names with various suffixes had to go through a complicated
sequence of tests that effectively partition the extensions into
several classes, with all extensions in a class being treated the same
but different classes being treated differently. And since this general
translation process is quite convoluted (which is not helped by it
being spread across several predicates), it is very hard to construct
a correctness argument for it.
It would be better to represent the different classes of extensions
explicitly, in a du type, with each function symbol of that type
representing all the extensions in the corresponding class (in the sense
of the paragraph above). However, getting there in one diff would make
that diff far too hard to test and to review. So this first diff
starts by simply making extension a notag type.
The above is the first step in implementing one old XXX. This diff
fully implements another old XXX, which is to make the argument order
of several predicates friendly to higher order code.
Add infrastructure for profiling how often this code makes directories.
Delete an unused type.
Add comments outlining proposed future improvements.
compiler/analysis.file.m:
compiler/analysis.m:
compiler/compile_target_code.m:
compiler/du_type_layout.m:
compiler/elds_to_erlang.m:
compiler/export.m:
compiler/fact_table.m:
compiler/file_kind.m:
compiler/find_module.m:
compiler/generate_dep_d_files.m:
compiler/grab_modules.m:
compiler/llds_out_file.m:
compiler/make.build.m:
compiler/make.m:
compiler/make.module_dep_file.m:
compiler/make.module_target.m:
compiler/make.program_target.m:
compiler/make.util.m:
compiler/mercury_compile_front_end.m:
compiler/mercury_compile_llds_back_end.m:
compiler/mercury_compile_main.m:
compiler/mercury_compile_middle_passes.m:
compiler/mercury_compile_mlds_back_end.m:
compiler/mlds_to_c_file.m:
compiler/mlds_to_cs_file.m:
compiler/mlds_to_java_file.m:
compiler/mmc_analysis.m:
compiler/mode_constraints.m:
compiler/module_cmds.m:
compiler/module_imports.m:
compiler/parse_tree_out.m:
compiler/prog_foreign.m:
compiler/read_modules.m:
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
compiler/write_deps_file.m:
compiler/write_module_interface_files.m:
compiler/xml_documentation.m:
Conform to the change to file_names.m.
Consistently use "Ext" for the abstract representation of extensions
and "ExtStr" for their string representation.
In a few places, add "XXX EXT" where the code manipulates extensions
as strings in a way that potentially inferferes with the partition
of extensions into classes.
In a few places, rename predicates to avoid ambiguities. factor out
common code, delete unneeded arguments, replace bools with bespoke types,
and make similar minor improvements.
In a few places, remove rafe-isms, such as the use ^elem.
compiler/grab_modules.m:
We used to scan the contents of each .opt file for dependencies twice
(with default option settings): once when deciding what .opt files
to read, and once when deciding what .int files to read to make sense
of those .opt files. Simplify and optimize this: avoid the second scan
by remembering the results of the first scan.
compiler/get_dependencies.m:
Since the first scan is all that is left, and it scans each .opt file
individually, compute the dependencies of each .opt file by itself.
Return the modules that a .opt file imports explicitly from the ones
it imports implicitly, in order to allow the latter to be represented
as a set of flags. ORing the flags of different .opt files together,
and then converting the resulting flags to a set of modules should be
faster than converting each .opt files' flags to a set of modules,
and then unioning those sets. Provide a predicate to OR the flags.
compiler/get_dependencies.m:
The interface of the predicates that computed implicit dependencies
used to return
(a) the set of modules a piece of code needs to ":- import_module", and
(b) the set of modules a piece of code needs to ":- use_module".
However, set (a) *always* contained only one module: builtin.m.
This lead to a lot of unnecessary operations on this "set".
The problem with those operations was not the performance impact,
which is negligible, but the fact that they made the code harder
to understand.
This diff therefore changes that interface to return only (b),
and to document what callers should use for (a).
compiler/grab_modules.m:
compiler/hlds_module.m:
compiler/module_imports.m:
Conform to the change above.
compiler/module_imports.m:
Delete the fields of the module_imports data structure that hold
the name of the module we are processing and its context, since
that info is available in the parse tree of the module's source code.
compiler/prog_item.m:
To make the above possible, fill in the context field of
just initialized, not-yet-filled in source code parse trees.
compiler/grab_modules.m:
Conform to the changes above.
