The option is disabled by default.
configure.ac:
Add --enable-lto option to the configure script to enable LTO/LTCG.
scripts/mgnuc.in:
scripts/ml.in:
Pass LTO options to the C compiler/linker.
compiler/compile_target_code.m:
compiler/options.m:
scripts/Mercury.config.in:
Add internal options to specify C compiler and linker LTO/LTCG flags.
library/string.m:
Add missing word.
Just write "code points" instead of "character" followed by
clarification in a few spots.
Delete _underscores_ which aren't particularly helpful.
library/string.m:
Define behaviour of set_char, det_set_char and unsafe_set_char on
ill-formed sequences. Also define them to throw an exception on an
attempt to set a null character or surrogate code point in a UTF-8
string.
Delete claim that unsafe_set_char is constant time. That would only
be true for the destructive mode of unsafe_set_char, and that mode
has been disabled for a long time.
Implement the defined behaviour for C and C# versions of
unsafe_set_char. The Java version already behaved as defined.
Use unsafe_set_char to implement set_char instead of duplicating
foreign code.
Replace a couple of uses of strcpy with MR_memcpy as it was
convenient to do so. (On OpenBSD, the linker issues a warning
whenever strcpy is used. Avoiding the warning is not high priority
but we might still like to eliminate all uses of strcpy eventually.)
tests/hard_coded/Mmakefile:
tests/hard_coded/string_set_char_ilseq.exp:
tests/hard_coded/string_set_char_ilseq.exp2:
tests/hard_coded/string_set_char_ilseq.m:
Add test case.
compiler/module_imports.m:
Replace the recomp_need_qualifier type with the recomp_avail type,
which contains more information about how a module became available
to the module being compiled. Specifically, besides recording whether
it was the subject of an import_module or use_module declaration
(which governes whether references must be fully qualified or not),
it says in which section that declaration was (i.e. where references
may appear at all), whether there was a use_module in the interface
and an import_module in the implementation section (in which case
references must be fully qualified in the interface section but not
in the implementation section).
Update the discussion of module timestamps to reflect our current
understanding of it.
compiler/grab_modules.m:
Indirectly fill in the affected slot of module timestamps with
more detailed info.
compiler/recompilation.usage.m:
Record the newly available information about module timestamps.
compiler/recompilation.check.m:
Read in the newly available information about module timestamps.
For the time being, accept both the old and new ways of recording
such information.
Update the code that makes decisions based on this information.
The new code should always make the same decisions as the old code,
but those decisions look to be appropriate in only *some* cases,
mark each such code fragment with XXX RECOMP401. (The 401 part
is there because, as discussed on m-rev, one probable bug resembles
Mantis bug #401.)
compiler/module_imports.m:
The module_timestamp type has long had a field of the need_qualifier type.
For a year or more now, this field has had a big comment on it explaining
why its use here does not match the semantics of the need_qualifier type.
This diff gives this field a new, bespoke type, that is isomorphic
to need_qualifier, but is completely deparate. Document the reason why
I *can't* document the semantics of this new type :-(
compiler/grab_modules.m:
Do the same kind of replacement on values that are used only to set
this field of module_timestamps.
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
Use the values of the new type to make the same decisions we used to make
using values of the need_qualifier type.
tests/recompilation/*.m*:
Bring the programming style of these modules up to date.
tests/recompilation/*.err_exp.2:
Expect the updated line numbers in messages.
Delete the need_qualifier field from timestamps.
library/string.m:
Use unsafe_append_string_pieces in Mercury implementations of
append_list and join_list. This has no practical effect as we have
foreign code implementations of both, for all target languages.
library/string.m:
Implement string.replace_all using unsafe_append_string_pieces to
avoid intermediate strings. Use unsafe_sub_string_search_start to
avoid repeated range checks.
library/string.m:
Delete long-deprecated substring/3 function and substring/4 predicate.
The newly introduced `string_piece' type has a substring/3 data
constructor which takes (start, end) offsets into the base string,
whereas the function and predicate take (start, count) arguments.
To reduce potential confusion, delete the deprecated function and
predicate.
Delete other deprecated substring predicates and functions as well.
tests/general/Mercury.options:
tests/general/string_foldl_substring.exp:
tests/general/string_foldl_substring.m:
tests/general/string_foldr_substring.exp:
tests/general/string_foldr_substring.m:
tests/hard_coded/Mercury.options:
tests/hard_coded/string_substring.m:
Delete tests for deprecated predicates.
tests/tabling/mercury_java_parser_dead_proc_elim_bug.m:
tests/tabling/mercury_java_parser_dead_proc_elim_bug2.m:
tests/valid/mercury_java_parser_follow_code_bug.m:
Replace calls to unsafe_substring with unsafe_between.
NEWS:
Announce the changes.
library/string.m:
Define behaviour of string.replace_all on ill-formed code unit
sequences when the pattern is empty.
Implement that behaviour.
Use better variable names in documentation of string.replace and
string.replace_all.
tests/general/string_replace.exp:
tests/general/string_replace.exp2:
tests/general/string_replace.m:
Extend test case.
Update code style.
compiler/prog_item.m:
Fix a bug: provide a slot for foreign_procs in parse_tree_plain_opts.
Eliminate unnecessary differences between parse_tree_{plain,trans}_opt
and parse_tree_int[0123].
compiler/parse_module.m:
As a temporary measure for testing,
- convert every .opt and .trans_opt file read in from its generic
parse_tree_opt representation to its specific parse_tree_plain_opt
or parse_tree_trans_opt representation, in order to check for
items that should not occur in the purpose-specific representations,
- and then convert it back.
Provide (not yet used) predicates for reading in optimization files
into their specific parse tree formats. These do the first conversion
but not the second.
compiler/parse_tree_out.m:
Add predicates for writing out parse_tree_{plain,trans}_opt directly,
without conversion to a generic opt file.
compiler/convert_parse_tree.m:
Make the above possible by adding conversions to and from
parse_tree_{plain,trans}_opt.
Give exported predicates more expressive names.
compiler/comp_unit_interface.m:
compiler/intermod.m:
Conform to the changes above.
compiler/add_pred.m:
compiler/grab_modules.m:
Minor style improvements.
compiler/add_foreign_proc.m:
Fix the wording of an error message.
tests/invalid/fp_dup_bug.err_exp:
Expect the updated wording.
library/io.m:
Put the contents of the implementation section of io.m into
the same order as the interface section, after some minor reordering
in the interface section.
library/Mercury.options:
Don't specify --no-warn-inconsistent-pred-order-clauses for io.m
anymore.
compiler/options.m:
As above.
configure.ac:
Require a recent-enough compiler, one that does not generate
invalid-by-the-recently-changed-standard interfaces.
library/string.m:
Make generic implementations of string.to_upper and string.to_lower
preserve ill-formed sequences. (The foreign language implementations
already did so.)
library/string.m:
Document that count_utf8_code_units throws an exception if the
string contains an unpaired surrogate code point.
Make the exception message thrown more useful to callers.
Delete unnecessary foreign_procs.
library/string.m:
Document that from_code_unit_list fails if the result string would
contain a null character, and enforce that in the Java and C#
implementations. It was already enforced in the C implementation.
Make from_code_unit_list fail if the code unit list contains an
invalid value (negative or >0xff or >0xffff).
Document that from_utf{8,16}_code_unit_list fails if the result
string would contain a null character.
Make from_utf8_code_unit_list call semidet_from_rev_char_list rather
than from_rev_char_list so that it fails as documented instead of
throwing an exception if the code unit list correctly encodes a list
of code points, but the code points cannot be encoded into a string.
Similarly for from_utf16_code_unit_list.
tests/hard_coded/Mmakefile:
tests/hard_coded/string_from_code_unit_list.exp:
tests/hard_coded/string_from_code_unit_list.exp2:
tests/hard_coded/string_from_code_unit_list.m:
Add test case.
extras/posix/posix.strerror.m:
Fix use of MR_strerror (it may return a pointer to a static string
instead of modifying the provided buffer).
Allocate string such that it will be attributed to the predicate
when profiling memory retention.
library/string.m:
Define sub_string_search_start to fail if the BeginAt parameter is
negative or past the end of the string to search. The original C
implementation did not check for an out-of-range starting offset,
and could crash the program. The C implementation was later amended
to fail instead, but not other implementations.
Check for negative starting offset in non-C implementations of
sub_string_search_start.
tests/hard_coded/string_sub_string_search.m:
Extend test case.
library/string.m:
Make C# implementation of sub_string_search perform ordinal
(Unicode code point) based string search, instead of a
culture-sensitive search.
library/string.m:
Make split_at_separator never consider ill-formed sequences in UTF-8
strings as potential separators, as they cannot contain any code
points that could satify any given DelimP predicate on code points.
Previously, split_at_separator would call DelimP(U+FFFD) for every
code unit in an ill-formed sequence.
library/string.m:
Make words_separator never consider ill-formed sequences in UTF-8
strings as potential separators, as they cannot contain any code
points that could satisfy any given SepP predicate on code points.
Previously, words_separator would call SepP(U+FFFD) for every code
unit in an ill-formed sequence.
library/string.m:
Make all_match(Pred, String) always fail if the string contains an
ill-formed code unit sequence, and strings use UTF-8 encoding.
Such sequences do not contain any code points that could satisfy a
test on code points. Previously, all_match would call Pred(U+FFFD)
for every code unit in an ill-formed sequence.
Define all_match to rule out an interpretation that could ignore
ill-formed sequences.
library/string.m:
Make prefix_length and suffix_length stop at an ill-formed sequence
in UTF-8 strings. Such a sequence does not contain any code point
that could satisfy a test on code points. Previously, prefix_length
and suffix_length would would call Pred(U+FFFD) for every code unit
in an ill-formed sequence.
Tweak documentation.
Delete obsolete comments.
library/string.m:
Fix C implementation of contains_char to fail when asked to test for
a surrogate code point in a string. It previously would (always)
succeed, which is a bug.
Fix generic implementation so that contains_char(String, '\uFFFD')
will not succeed just because String contains an ill-formed sequence
(in UTF-8 grades).
Delete obsolete comment.
library/string.m:
Add index_next_repl, unsafe_index_next_repl, prev_index_repl,
unsafe_prev_index_repl predicates. These are internal for now,
so we can try them out in the string module without committing
to the interface.
library/string.m:
Fix definition of index/3 and index_next/4 to account for an offset
into a non-initial code unit in a well-formed code unit sequence.
Similarly for prev_index/4.
compiler/convert_interface.m:
As above. I am about to add code to convert optimization files as well.
compiler/parse_tree.m:
Include the module under its new name.
compiler/notes/compiler_design.html:
Document the module under its new name.
compiler/comp_unit_interface.m:
compiler/intermod.m:
compiler/parse_module.m:
Import the module under its new name.
