Merge namespace scopes in eval.c

The big deletion patch left some identical namespace scopes sitting
next to one another.  This patch removes the redundant "namespace
expr"s.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* eval.c: Merge "namespace" scopes.

Inline expr_builder methods

This inlines the expr_builder constructor and release method.  These
are straightforward, so this seemed simpler.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* parser-defs.h (struct expr_builder) <expr_builder>: Inline.
<release>: Inline.
* parse.c (expr_builder::expr_builder, expr_builder::release):
Remove.

Inline expression constructor

The struct expression constructor no longer does any real work, so
this inlines it.  The default destructor can also be used now as well.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* parse.c (expression::expression, expression::~expression):
Remove.
* expression.h (struct expression): Inline constructor.  Remove
destructor.

Remove BINOP_END

BINOP_END is used only as a "meaningless" value in various tables.
This patch changes these to use OP_NULL instead, and removes
BINOP_END.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* std-operator.def (BINOP_END): Remove.
* p-exp.y (tokentab3, tokentab2): Use OP_NULL, not BINOP_END.
* go-exp.y (tokentab2): Use OP_NULL, not BINOP_END.
* f-exp.y (dot_ops, f77_keywords): Use OP_NULL, not BINOP_END.
* d-exp.y (tokentab2, ident_tokens): Use OP_NULL, not BINOP_END.
* c-exp.y (tokentab3, tokentab2, ident_tokens): Use OP_NULL, not
BINOP_END.

Remove OP_UNUSED_LAST

OP_UNUSED_LAST is no longer needed with C++ -- the trailing comma is
fine.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expression.h (enum exp_opcode) <OP_UNUSED_LAST>: Remove.

Remove OP_EXTENDED0

OP_EXTENDED0 was only used for an assertion in the code to rewrite an
expression into prefix form.  That code is gone, so this patch removes
the constant.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* std-operator.def (OP_EXTENDED0): Remove.

Remove unused Ada opcodes

Several Ada expression opcodes are now unused, and can be removed.
Most of these are handled in a different way by the code.
OP_ATR_IMAGE, however, was never implemented.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* std-operator.def (OP_NAME, OP_ATR_IMAGE, OP_ATR_MODULUS)
(OP_OTHERS, OP_CHOICES, OP_POSITIONAL, OP_DISCRETE_RANGE):
Remove.

Remove unused Modula-2 opcodes

As noted in an earlier patch, Modula-2 defined some opcodes but then
never implemented them.  This patch removes the unnecessary constants.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* std-operator.def (UNOP_CAP, UNOP_CHR, UNOP_ORD, UNOP_FLOAT)
(UNOP_MAX, UNOP_MIN, UNOP_ODD, UNOP_TRUNC, OP_M2_STRING): Remove.

Remove two Ada opcodes

The OP_ATR_MIN and OP_ATR_MAX constants aren't truly needed.
Internally, they are converted to BINOP_MIN and BINOP_MAX.  This patch
removes them in favor of simple reuse.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* std-operator.def (OP_ATR_MIN, OP_ATR_MAX): Remove.
* ada-lang.c (ada_binop_minmax): Update.
* ada-exp.h (ada_binop_min_operation, ada_binop_max_operation):
Use BINOP_MIN and BINOP_MAX.

Remove union exp_element

This removes union exp_element functions that either create such
elements or walk them.  struct expression no longer holds
exp_elements.  A couple of language_defn methods are also removed, as
they are obsolete.

Note that this patch also removes the print_expression code.  The only
in-tree caller of this was from dump_prefix_expression, which is only
called when expression debugging is enabled.  Implementing this would
involve a fair amount of code, and it seems to me that prefix dumping
is preferable anyway, as it is unambiguous.  So, I have not
reimplemented this feature.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* value.h (evaluate_subexp_with_coercion): Don't declare.
* parse.c (exp_descriptor_standard): Remove.
(expr_builder::expr_builder, expr_builder::release): Update.
(expression::expression): Remove size_t parameter.
(expression::~expression): Simplify.
(expression::resize): Remove.
(write_exp_elt, write_exp_elt_opcode, write_exp_elt_sym)
(write_exp_elt_msym, write_exp_elt_block, write_exp_elt_objfile)
(write_exp_elt_longcst, write_exp_elt_floatcst)
(write_exp_elt_type, write_exp_elt_intern, write_exp_string)
(write_exp_string_vector, write_exp_bitstring): Remove.
* p-lang.h (class pascal_language) <opcode_print_table,
op_print_tab>: Remove.
* p-lang.c (pascal_language::op_print_tab): Remove.
* opencl-lang.c (class opencl_language) <opcode_print_table>:
Remove.
* objc-lang.c (objc_op_print_tab): Remove.
(class objc_language) <opcode_print_table>: Remove.
* m2-lang.h (class m2_language) <opcode_print_table,
op_print_tab>: Remove.
* m2-lang.c (m2_language::op_print_tab): Remove.
* language.h (struct language_defn) <post_parser, expression_ops,
opcode_print_table>: Remove.
* language.c (language_defn::expression_ops)
(auto_or_unknown_language::opcode_print_table): Remove.
* go-lang.h (class go_language) <opcode_print_table,
op_print_tab>: Remove.
* go-lang.c (go_language::op_print_tab): Remove.
* f-lang.h (class f_language) <opcode_print_table>: Remove
<op_print_tab>: Remove.
* f-lang.c (f_language::op_print_tab): Remove.
* expression.h (union exp_element): Remove.
(struct expression): Remove size_t parameter from constructor.
<resize>: Remove.
<first_opcode>: Update.
<nelts, elts>: Remove.
(EXP_ELEM_TO_BYTES, BYTES_TO_EXP_ELEM): Remove.
(evaluate_subexp_standard, print_expression, op_string)
(dump_raw_expression): Don't declare.
* expprint.c (print_expression, print_subexp)
(print_subexp_funcall, print_subexp_standard, op_string)
(dump_raw_expression, dump_subexp, dump_subexp_body)
(dump_subexp_body_funcall, dump_subexp_body_standard): Remove.
(dump_prefix_expression): Update.
* eval.c (evaluate_subexp): Remove.
(evaluate_expression, evaluate_type): Update.
(evaluate_subexpression_type): Remove.
(fetch_subexp_value): Remove "pc" parameter.  Update.
(extract_field_op, evaluate_struct_tuple, evaluate_funcall)
(evaluate_subexp_standard, evaluate_subexp_for_address)
(evaluate_subexp_with_coercion, evaluate_subexp_for_sizeof)
(evaluate_subexp_for_cast): Remove.
(parse_and_eval_type): Update.
* dtrace-probe.c (dtrace_probe::compile_to_ax): Update.
* d-lang.c (d_op_print_tab): Remove.
(class d_language) <opcode_print_table>: Remove.
* c-lang.h (c_op_print_tab): Don't declare.
* c-lang.c (c_op_print_tab): Remove.
(class c_language, class cplus_language, class asm_language, class
minimal_language) <opcode_print_table>: Remove.
* breakpoint.c (update_watchpoint, watchpoint_check)
(watchpoint_exp_is_const, watch_command_1): Update.
* ax-gdb.h (union exp_element): Don't declare.
* ax-gdb.c (const_var_ref, const_expr, maybe_const_expr)
(gen_repeat, gen_sizeof, gen_expr_for_cast, gen_expr)
(gen_expr_binop_rest): Remove.
(gen_trace_for_expr, gen_eval_for_expr, gen_printf): Update.
* ada-lang.c (ada_op_print_tab): Remove.
(class ada_language) <post_parser, opcode_print_table>: Remove.

Remove now-unused C evaluator code

Now that the C parser has switched to the new style, there is no need
for the old C evaluation code.  This affects some other languages that
were relying on the C code.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* go-lang.c (go_language::expression_ops): Don't declare.
* go-lang.h (class go_language) <expression_ops>: Remove.
* opencl-lang.c (evaluate_subexp_opencl, exp_descriptor_opencl):
Remove.
(class opencl_language) <expression_ops>: Remove.
* d-lang.c (class d_language) <expression_ops>: Remove.
* c-lang.h (evaluate_subexp_c, exp_descriptor_c): Don't declare.
* c-lang.c (evaluate_subexp_c, exp_descriptor_c): Remove.
(class c_language, class cplus_language, class asm_language)
(class minimal_language) <expression_ops>: Remove.

Remove now-unused Ada evaluator code

Now that the Ada parser has switched to the new style, there is no
need for the old Ada evaluation code.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (resolve_subexp, replace_operator_with_call)
(evaluate_subexp_type, assign_aggregate)
(aggregate_assign_positional, aggregate_assign_from_choices)
(aggregate_assign_others, ada_evaluate_subexp_for_cast)
(ada_evaluate_subexp, ADA_OPERATORS, ada_operator_length)
(ada_operator_check, ada_forward_operator_length)
(ada_dump_subexp_body, ada_print_subexp, ada_exp_descriptor):
Remove.
(post_parser): Update.
(class ada_language) <expresssion_ops>: Remove.

Remove now-unused Modula-2 evaluator code

Now that the Modula-2 parser has switched to the new style, there is
no need for the old Modula-2 evaluation code.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* m2-lang.h (class m2_language) <expresssion_ops,
exp_descriptor_modula2>: Remove.
* m2-lang.c (evaluate_subexp_modula2)
(m2_language::exp_descriptor_modula2): Remove.

Remove now-unused Fortran evaluator code

Now that the Fortran parser has switched to the new style, there is no
need for the old Fortran evaluation code.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-lang.h (class f_language) <expresssion_ops>: Remove.
<exp_descriptor_tab>: Remove.
* f-lang.c (fortran_value_subarray, evaluate_subexp_f)
(operator_length_f, print_unop_subexp_f, print_binop_subexp_f)
(print_subexp_f, dump_subexp_body_f, operator_check_f)
(f_language::exp_descriptor_tab, fortran_prepare_argument):
Remove.

Remove now-unused Rust evaluator code

Now that the Rust parser has switched to the new style, there is no
need for the old Rust evaluation code.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.h (class rust_language) <expression_ops,
exp_descriptor_tab>: Remove.
* rust-lang.c (rust_evaluate_funcall): Remove.
(rust_range, rust_subscript, eval_op_rust_complement): Don't use
EVAL_SKIP.
(rust_evaluate_subexp): Remove.
(rust_aggregate_operation::evaluate): Don't use EVAL_SKIP.
(rust_operator_length, rust_dump_subexp_body, rust_print_subexp)
(rust_operator_check, rust_language::exp_descriptor_tab): Remove.

Convert ada-exp.y to use operations

This converts the Ada parser to generate operations rather than
exp_elements.

This was the most difficult of the parser conversions, partly due to
the decision to integrate Ada expression resolution into the parse,
and partly due to Ada aggregregate assignment.  A couple of new
per-parse globals are introduced, along with a number of helper
functions.  Resolution is done in 'ada_pop', yielding the unfortunate
rule that ada-exp.y should generally not use parser_state::pop
(exceptions are marked).

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-exp.y: Create operations.
(empty_stoken): Remove.
(ada_pop, ada_wrap, ada_addrof, ada_un_wrap2, ada_wrap2)
(ada_wrap_op, ada_wrap3, ada_funcall): New functions.
(components): New global.
(push_component, choice_component, pop_component, pop_components):
New functions.
(associations): New global
(push_association, pop_association, pop_associations): New
functions.
(ada_parse): Update.
(write_var_from_sym, write_int): Create operations.
(write_exp_op_with_string): Remove.
(write_object_renaming, write_selectors, write_ambiguous_var)
(write_var_or_type, write_name_assoc): Create operations.
* ada-lang.h (ada_index_type): Declare.
* ada-lang.c (ada_index_type): No longer static.

Convert f-exp.y to use operations

This converts the Fortran parser to generate operations rather than
exp_elements.  A couple of tests of expression debug dumping are
updated to follow the new output.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-exp.y: Create operations.
(f_language::parser): Update.

gdb/testsuite/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* gdb.fortran/debug-expr.exp: Update tests.

Convert m2-exp.y to use operations

This converts the Modula-2 parser to generate operations rather than
exp_elements.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* m2-exp.y: Create operations.
(m2_language::parser): Update.

Convert p-exp.y to use operations

This converts the Pascal parser to generate operations rather than
exp_elements.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* p-exp.y: Create operations.
(pascal_language::parser): Update.

Convert d-exp.y to use operations

This converts the D parser to generate operations rather than
exp_elements.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* d-exp.y: Create operations.
(d_parse): Update.

Convert go-exp.y to use operations

This converts the Go parser to generate operations rather than
exp_elements.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* go-exp.y: Create operations.
(go_language::parser): Update.

Convert c-exp.y to use operations

This converts the C parser to generate operations rather than
exp_elements.

One test needed a tweak in order to handle the different debugging
output.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* objc-lang.c (end_msglist): Create operations.
* c-exp.y: Change parser to create operations.
(write_destructor_name): Remove.
(c_parse): Update.

gdb/testsuite/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* gdb.base/debug-expr.exp: Update expected dump output.

Convert rust-exp.y to use operations

This converts the Rust parser to generate operations rather than
exp_elements.

The Rust parser already made its own AST, that it then lowered to GDB
expressions.  Ironically, this made conversion trickier, rather than
simpler, than the other parsers, primarily due to the way that binary
operations were lowered.  Perhaps in the future, converting the Rust
parser to directly create operations while parsing would be
worthwhile.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-exp.y: Create operations.
(rust_parser::convert_params_to_expression): Change return type.
(binop_maker_ftype): New typedef.
(maker_map): New global.
(rust_parser::convert_ast_to_expression): Change return type.
(rust_language::parser): Update.
(_initialize_rust_exp): Initialize maker_map.

Convert stap probes to create operations

This changes the stap probe code to create operations, rather than
exp_elements.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* stap-probe.c (binop_maker_ftype): New typedef.
(stap_maker_map): New global.
(stap_make_binop): New function.
(stap_parse_register_operand): Return operation_up.
(stap_parse_single_operand, stap_parse_argument_conditionally)
(stap_parse_argument_1): Likewise.
(stap_parse_argument): Create operations.
(stap_probe::parse_arguments): Update.
(_initialize_stap_probe): Initialize stap_maker_map.
* ppc-linux-tdep.c (ppc_stap_parse_special_token): Change return
type.
* i386-tdep.h (i386_stap_parse_special_token): Change return
type.
* i386-tdep.c (i386_stap_parse_special_token_triplet)
(i386_stap_parse_special_token_three_arg_disp)
(i386_stap_parse_special_token): Change return type.
* gdbarch.sh (stap_parse_special_token): Change return type.
* gdbarch.c: Rebuild.
* gdbarch.h: Rebuild.
* arm-linux-tdep.c (arm_stap_parse_special_token): Change return
type.
* aarch64-linux-tdep.c (aarch64_stap_parse_special_token): Change
return type.

Convert dtrace probes to use operations

This changes dtrace to use the new operation type.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* gdbarch.sh (dtrace_parse_probe_argument): Change return type.
* gdbarch.h: Rebuild.
* gdbarch.c: Rebuild.
* dtrace-probe.c (dtrace_probe::build_arg_exprs): Update.
* amd64-linux-tdep.c (amd64_dtrace_parse_probe_argument): Change
return type.
(amd64_dtrace_parse_probe_argument): Update.

Add operation-related methods to parser_state

This adds several operation-related methods to parser_state.  These
methods make it more convenient to change the parsers to be
operation-based.

Because byacc has poor support for C++, a stack of operations is added
to parser_state.  A parser can push operations, then later pop them
for combination into new operations.  This approach avoids the memory
leaks that would result if raw pointers were used in the parsers, at
the cost of parser productions not being type-safe (they can't
indicate that they return an operation).

This also introduces analogs of some write_exp functions, like
write_exp_string_vector, write_dollar_variable, and
write_exp_symbol_reference.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* parser-defs.h (struct parser_state) <push, push_new,
push_c_string, push_symbol, push_dollar, pop, pop_vector, wrap,
wrap2>: New methods.
<m_operations>: New member.
* parse.c (parser_state::push_c_string)
(parser_state::push_symbol, parser_state::push_dollar): New
methods.

Add completion for operations

This patch adds the necessary support for field name completion for
expressions using class operation.

This patch takes an approach similar to what is done today.  It might
be good, in the future, to change completion to be a method on the
base class, to enable context-sensitive completion in more areas.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* parser-defs.h (struct expr_completion_state) <expout_last_op>:
New member.
(struct parser_state) <mark_struct_expression>: New method.
* parse.c (parser_state::mark_struct_expression): Update assert.
(parser_state::mark_struct_expression): New method.
(parser_state::mark_completion_tag): Update assert.
(parse_expression_for_completion): Handle expout_last_op.

Add an expr::operation_up to struct expression

This adds an expr::operation_up to struct expression, and then
modifies various parts of GDB to use this member when it is non-null.
The list of such spots was a bit surprising to me, and found only
after writing most of the code and then noticing what no longer
compiled.

In a few spots, new accessor methods are added to operation
subclasses, so that code that dissects an expression will work with
the new scheme.

After this change, code that constructs an expression can be switched
to the new form without breaking.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-exp.h (class ada_var_value_operation) <get_symbol>: Remove;
now in superclass.
* value.h (fetch_subexp_value): Add "op" parameter.
* value.c (init_if_undefined_command): Update.
* tracepoint.c (validate_actionline, encode_actions_1): Update.
* stap-probe.c (stap_probe::compile_to_ax): Update.
* printcmd.c (set_command): Update.
* ppc-linux-nat.c (ppc_linux_nat_target::check_condition):
Update.
* parser-defs.h (struct expr_builder) <set_operation>: New
method.
* parse.c (parse_exp_in_context, exp_uses_objfile): Update.
* expression.h (struct expression) <first_opcode>: Update.
<op>: New member.
* expprint.c (dump_raw_expression, dump_prefix_expression):
Update.
* expop.h (class var_value_operation) <get_symbol>: New method.
(class register_operation) <get_name>: New method.
(class equal_operation): No longer a typedef, now a subclass.
(class unop_memval_operation) <get_type>: New method.
(class assign_operation) <get_lhs>: New method.
(class unop_cast_operation) <get_type>: New method.
* eval.c (evaluate_expression, evaluate_type)
(evaluate_subexpression_type): Update.
(fetch_subexp_value): Add "op" parameter.
(parse_and_eval_type): Update.
* dtrace-probe.c (dtrace_probe::compile_to_ax): Update.
* breakpoint.c (update_watchpoint, watchpoint_check)
(watchpoint_exp_is_const, watch_command_1): Update.
* ax-gdb.c (gen_trace_for_expr, gen_eval_for_expr, gen_printf):
Update.

Remove use of op_string

After switching to the new expression implementation, there will no
need for op_string.  Before deleting it, the one call outside of the
expression-printing code must be removed.  That is what this patch
does.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_value_binop): Do not use op_string.

Implement Ada assignment

Assignment is the most complicated Ada expression, because
implementing aggregate assignment involves several specialized
opcodes.

This patch does this implementation by introducing new abstract
classes that are used to represent the various parts of aggregate
assignment.  This makes the code somewhat cleaner, and, by avoiding
the over-use of 'operation' subclasses, avoids the need for dissection
using dynamic_cast (though a few are still needed here).

I believe this patch fixes a latent bug in the handling of
aggregate_assign_from_choices.  That code does:

      if (op == OP_DISCRETE_RANGE)
{
  choice_pos += 1;
  lower = value_as_long (ada_evaluate_subexp (NULL, exp, pos,
      EVAL_NORMAL));
  upper = value_as_long (ada_evaluate_subexp (NULL, exp, pos,
      EVAL_NORMAL));
}

However, I think 'choice_pos' should be used in the calls, rather than
'pos'.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expprint.c (dump_for_expression): New overload.
* expop.h (check_objfile, dump_for_expression): Declare new
overloads.
* ada-lang.c (check_objfile): New overload.
(assign_component, ada_aggregate_component::uses_objfile)
(ada_aggregate_component::dump, ada_aggregate_component::assign)
(ada_aggregate_component::assign_aggregate)
(ada_positional_component::uses_objfile)
(ada_positional_component::dump, ada_positional_component::assign)
(ada_discrete_range_association::uses_objfile)
(ada_discrete_range_association::dump)
(ada_discrete_range_association::assign)
(ada_name_association::uses_objfile, ada_name_association::dump)
(ada_name_association::assign)
(ada_choices_component::uses_objfile, ada_choices_component::dump)
(ada_choices_component::assign)
(ada_others_component::uses_objfile, ada_others_component::dump)
(ada_others_component::assign, ada_assign_operation::evaluate):
New methods.
* ada-exp.h (ada_string_operation) <get_name>: New method.
(class ada_assign_operation): New.
(class ada_component): New.
(ada_component_up): New typedef.
(class ada_aggregate_operation, class ada_aggregate_component)
(class ada_positional_component, class ada_others_component)
(class ada_association): New.
(ada_association_up): New typedef.
(class ada_choices_component)
(class ada_discrete_range_association)
(class ada_name_association): New.

Implement Ada resolution

Ada has a parser post-pass that implements "resolution".  This process
replaces some opcodes with function calls.  For example, a "+"
operation might be replaced with a call to the appropriate overloaded
function.

This differs from the approach taken for the same problem in C++.
However, in this series I chose not to try to make changes outside of
rewrite the expression data structure.  So, resolution remains.

The new approach to resolution is to introduce an interface class,
that some concrete operations implement.  Then, the Ada code will use
this to resolve the expression tree.  Because new-style expressions
are built as ordinary objects, and don't require rewriting the data
structure in place, in the new code this processing will be done in
the parser.  By the end of the series, some special cases in this area
that exist only for Ada will be removed.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_var_value_operation::resolve)
(ada_funcall_operation::resolve)
(ada_ternop_slice_operation::resolve): New methods.
* ada-exp.h (struct ada_resolvable): New.
(class ada_var_value_operation): Derive from ada_resolvable.
<get_block, resolve>: New methods.
(class ada_funcall_operation): Derive from ada_resolvable.
<resolve>: New method.
(class ada_ternop_slice_operation): Derive from ada_resolvable.
<resolve>: New method.

Implement function calls for Ada

This implements function calls for Ada.  This takes a different
approach than that used for other languages, primarily because Ada
requires special treatment generally.  The "ordinary" special case for
just the callee didn't really apply neatly here; there's only one case
in Ada needing special callee treatment.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_funcall_operation::evaluate): New method.
* ada-exp.h (class ada_var_msym_value_operation) <get_symbol>: New
method.
(class ada_funcall_operation): New.

Introduce ada_structop_operation

This adds class ada_structop_operation, which implements
STRUCTOP_STRUCT for Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_structop_operation::evaluate): New method.
* ada-exp.h (class ada_structop_operation): New.

Introduce ada_unop_ind_operation

This adds class ada_unop_ind_operation, which implements UNOP_IND for
Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_unop_ind_operation::evaluate): New method.
* ada-exp.h (class ada_unop_ind_operation): New.

Introduce ada_binop_exp_operation

This adds class ada_binop_exp_operation, which implements BINOP_EXP
for Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_binop_exp): No longer static.
* ada-exp.h (ada_binop_exp_operation): New typedef.

Introduce ada_atr_val_operation

This adds class ada_atr_val_operation, which implements OP_ATR_VAL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_val_atr): No longer static.
(ada_atr_val_operation::evaluate): New method.
* ada-exp.h (class ada_atr_val_operation): New.

Introduce ada_pos_operation

This adds class ada_pos_operation, a new typedef that implements
OP_ATR_POS.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_pos_atr): No longer static.
* ada-exp.h (ada_pos_operation): New typedef.

Refactor value_pos_atr

This refactors value_pos_atr to be directly usable by a new operation
implementation.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_pos_atr): Rename from value_pos_atr.  Change
parameters.
(ada_evaluate_subexp): Use it.

Implement Ada min and max operations

This implement the Ada min and max operations using an existing
template class.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_binop_minmax): No longer static.
* ada-exp.h (ada_binop_min_operation, ada_binop_max_operation):
New typedefs.

Introduce ada_var_msym_value_operation

This adds class ada_var_msym_value_operation, which implements
OP_VAR_MSYM_VALUE for Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_var_msym_value_operation::evaluate_for_cast):
New method.
* ada-exp.h (class ada_var_msym_value_operation): New.

Introduce ada_var_value_operation

This adds class ada_var_value_operation, which implements OP_VAR_VALUE
for Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_var_value_operation::evaluate_for_cast)
(ada_var_value_operation::evaluate): New methods.
* ada-exp.h (class ada_var_value_operation): New.

Implement some Ada OP_ATR_ operations

This implements a few Ada OP_ATR_ operations.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_unop_atr_operation::evaluate): New method.
* ada-exp.h (class ada_unop_atr_operation): New.

Introduce ada_binop_in_bounds

This adds class ada_binop_in_bounds, which implements BINOP_IN_BOUNDS.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_binop_in_bounds): No longer static.
* ada-exp.h (class ada_binop_in_bounds_operation): New.

Introduce ada_ternop_slice

This adds class ada_ternop_slice, which implements TERNOP_SLICE for
Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_ternop_slice): No longer static.
* ada-exp.h (class ada_ternop_slice_operation): New.

Introduce ada_bitwise_operation

This adds class ada_bitwise_operation, which is used to implement the
Ada bitwise operators.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-exp.h (ada_bitwise_operation): New template class.
(ada_bitwise_and_operation, ada_bitwise_ior_operation)
(ada_bitwise_xor_operation): New typedefs.

Implement Ada equality operators

This implements the Ada equal and not-equal operators.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_equal_binop): No longer static.
* ada-exp.h (class ada_binop_equal_operation): New.

Implement Ada multiplicative operators

This implements the Ada multiplicative operators, using an existing
template class.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_mult_binop): No longer static.
* ada-exp.h (ada_binop_mul_operation ada_binop_div_operation)
(ada_binop_rem_operation, ada_binop_mod_operation): New typedefs.

Introduce ada_binop_addsub_operation

This adds class ada_binop_addsub_operation, which implements the Ada +
and - operators.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_binop_addsub_operation::evaluate): New method.
* ada-exp.h (class ada_binop_addsub_operation): New.

Split out some Ada type resolution code

This splits some Ada type resolution code out of resolve_subexp into
new functions that can be reused.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.h (ada_find_operator_symbol, ada_resolve_funcall)
(ada_resolve_variable): Declare.
* ada-lang.c (ada_find_operator_symbol, ada_resolve_funcall)
(ada_resolve_variable): New functions.
(resolve_subexp): Update.

Implement OpenCL ternary conditional operator

This implements the ?: ternary conditional operator for OpenCL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_ternop_cond_operation::evaluate): New
method.
* c-exp.h (class opencl_ternop_cond_operation): New.

Implement OpenCL logical binary operations

This implements "&&" and "||" for OpenCL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_logical_binop_operation::evaluate): New
method.
* c-exp.h (class opencl_logical_binop_operation): New.

Introduce opencl_structop_operation

This adds class opencl_structop_operation, which implements
STRUCTOP_STRUCT for OpenCL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_structop_operation::evaluate): New
method.
* c-exp.h (class opencl_structop_operation): New.

Introduce opencl_notequal_operation

This adds the opencl_notequal_operation typedef, implementing "!=" for
OpenCL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_logical_not): No longer static.  Change
parameters.
(evaluate_subexp_opencl): Update.
* c-exp.h (opencl_notequal_operation): New typedef.

Implement OpenCL binary operations

This implements various OpenCL binary operations, adding a new
template class to do so.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_relop, eval_opencl_assign): No longer
static.  Change parameters.
(eval_opencl_assign): No longer static.  Add "op" parameter.
(evaluate_subexp_opencl): Update.
* c-exp.h (opencl_binop_operation): New template class.
(opencl_assign_operation, opencl_equal_operation)
(opencl_notequal_operation, opencl_less_operation)
(opencl_gtr_operation, opencl_geq_operation)
(opencl_leq_operation): New typedefs.

Introduce opencl_cast_type_operation

This adds class opencl_cast_type_operation, which implements
UNOP_CAST_TYPE for OpenCL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* opencl-lang.c (opencl_value_cast): No longer static.
* c-exp.h (opencl_cast_type_operation): New typedef.

Implement fortran_allocated_operation

This implements the Fortran ALLOCATED intrinsic.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-exp.h (eval_op_f_allocated): Declare.
(fortran_allocated_operation): New typedef.
* f-lang.c (eval_op_f_allocated): No longer static.

Implement Fortran associated operations

This implements the Fortran 1- and 2-argument "associated" operations.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-lang.c (eval_op_f_associated): New functions.
* f-exp.h (fortran_associated_1arg, fortran_associated_2arg): New
typedefs.

Introduce classes for Fortran bound intrinsics

This adds class fortran_bound_1arg and fortran_bound_2arg, to
implement the Fortran lbound and ubound intrinsics.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-lang.c (fortran_bound_1arg::evaluate)
(fortran_bound_2arg::evaluate): New methods.
* f-exp.h (class fortran_bound_1arg, class fortran_bound_2arg):
New.

Introduce fortran_undetermined

This adds class fortran_undetermined, which implements
OP_F77_UNDETERMINED_ARGLIST.  fortran_range_operation is also added
here, as it is needed by fortran_undetermined.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_addr_operation) <get_expression>: New
method.
* f-lang.c (fortran_undetermined::value_subarray)
(fortran_undetermined::evaluate): New methods.
(fortran_prepare_argument): New overload.
* f-exp.h (class fortran_range_operation)
(class fortran_undetermined): New classes.

Implement Rust funcall operation

This adds the special code needed to handle the Rust function call
operation.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (rust_structop::evaluate_funcall): New method.
* rust-exp.h (class rust_structop) <evaluate_funcall>: Declare
method.

Implement function call operations

This implement function call operations.

The current function call code relies on some very lengthy code
(evaluate_funcall is 398 lines...) to distinguish between the
different opcodes that might appear in the callee position.

Rather than try to replicate this, and have a function that tried to
dissect many different kinds of operation subclass, this patch instead
puts the work into the callee.  A new operation::evaluate_funcall
method is added, and then this is overridden in the classes that
require special treatment.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expression.h (class operation) <evaluate_funcall>: New methods.
* expop.h (class scope_operation) <evaluate_funcall>: New method.
(class var_value_operation) <evaluate_funcall>: New method.
(class structop_base_operation) <evaluate_funcall>: New method.
(class var_msym_value_operation) <evaluate_funcall>: New method.
(class structop_member_base): New class.
(class structop_member_operation): Derive from
structop_member_base.
(class structop_mptr_operation): Derive from
structop_member_base.
(class funcall_operation): New class.
* eval.c (operation::evaluate_funcall)
(var_value_operation::evaluate_funcall)
(scope_operation::evaluate_funcall)
(structop_member_base::evaluate_funcall)
(structop_base_operation::evaluate_funcall): New methods.

Introduce array_operation

This adds class array_operation, which implements OP_ARRAY.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class array_operation): New.
* eval.c (array_operation::evaluate_struct_tuple)
(array_operation::evaluate): New methods.

Introduce class adl_func_operation

This adds class adl_func_operation, which implements
argument-dependent lookup function calls.

Other function calls will be handled in a different way.  However,
because ADL calls were created in a single spot in the C++ parser, and
because they had different semantics from the other cases, it was
convenient to treat them specially.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class adl_func_operation): New.
* eval.c (adl_func_operation::evaluate): New method.

Introduce ada_unop_range_operation

This adds class ada_unop_range_operation, which implements
UNOP_IN_RANGE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_unop_in_range): No longer static.
* ada-exp.h (class ada_unop_range_operation): New.

Implement some Ada unary operations

This implements a few Ada unary operations, using the existing
unop_operation template class.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_unop_neg, ada_atr_tag, ada_atr_size, ada_abs):
No longer static.
* ada-exp.h (ada_neg_operation, ada_atr_tag_operation)
(ada_atr_size_operation, ada_abs_operation): New typedefs.

Implement the "&&" and "||" operators

This implements the "&&" and "||" operators.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class logical_and_operation)
(class logical_or_operation): New.
* eval.c (logical_and_operation::evaluate)
(logical_or_operation::evaluate): New methods.
* ax-gdb.c (logical_and_operation::do_generate_ax)
(logical_or_operation::do_generate_ax): New methods.

Add two simple Modula-2 operations

This adds a couple of simple Modula-2 operations.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* m2-lang.c (eval_op_m2_high, eval_op_m2_subscript): No longer
static.
* m2-exp.h: New file.

Introduce rust_aggregate_operation

This adds class rust_aggregate_operation, which implements
OP_AGGREGATE for Rust.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (rust_aggregate_operation::evaluate): New method.
* rust-exp.h (class rust_aggregate_operation): New.

Implement Rust field operations

This implements the field operations STRUCTOP_STRUCT and
STRUCTOP_ANONYMOUS, for Rust.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (eval_op_rust_struct_anon, eval_op_rust_structop):
No longer static.
* rust-exp.h (class rust_struct_anon): New.
(class rust_structop): New.

Introduce rust_range_operation

This adds class rust_range_operation, which implements OP_RANGE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (rust_range): No longer static.
* rust-exp.h (class rust_range_operation): New.

Introduce rust_subscript_operation

This adds class rust_subscript_operation, which implements
BINOP_SUBSCRIPT for Rust.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (rust_subscript): No longer static.
* rust-exp.h (class rust_subscript_operation): New.

Introduce rust_unop_ind_operation

This adds class rust_unop_ind_operation, which implements UNOP_IND for
Rust.  Rust requires a special case here to handle trait objects.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (eval_op_rust_ind): No longer static.  Add "opcode"
parameter.
(rust_evaluate_subexp): Update.
* rust-exp.h (class rust_unop_ind_operation): New.

Implement some Rust operations

This implements some straightforward Rust operations, using existing
template classes.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* rust-lang.c (eval_op_rust_complement, eval_op_rust_array): No
longer static.  Add "opcode" parameter.
(rust_evaluate_subexp): Update.
* rust-exp.h: New file.

Implement several Fortran operations

This implements several straightforward Fortran operations, primarily
by reusing existing template classes.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* f-lang.c (eval_op_f_abs, eval_op_f_mod, eval_op_f_ceil)
(eval_op_f_floor, eval_op_f_modulo, eval_op_f_cmplx)
(eval_op_f_kind): No longer static.  Add "opcode" parameter.
(evaluate_subexp_f): Update.
* f-exp.h: New file.

Introduce ada_ternop_range_operation

This adds class ada_ternop_range_operation, which implements
TERNOP_IN_RANGE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_ternop_range_operation::evaluate): New method.
* ada-exp.h (class ada_ternop_range_operation): New.

Introduce ada_qual_operation

This adds class ada_qual_operation, which implements UNOP_QUAL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_qual_operation::evaluate): New method.
* ada-exp.h (class ada_qual_operation): New.

Introduce ada_string_operation

This adds class ada_string_operation, which implements string
constants for Ada.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_string_operation::evaluate): New method.
* ada-exp.h (class ada_string_operation): New.

Introduce ada_wrapped_operation

This adds class ada_wrapped_operation, which is used to wrap some
generic operations with a bit of Ada-specific handling.  This
corresponds to the old "default" case in ada_evaluate_subexp.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ada-lang.c (ada_wrapped_operation::evaluate): New method.
* ada-exp.h: New file.

Introduce multi_subscript_operation

This adds class multi_subscript_operation, which implements
MULTI_SUBSCRIPT.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class multi_subscript_operation): New.
* eval.c (multi_subscript_operation::evaluate): New method.

Introduce objc_msgcall_operation

This adds class objc_msgcall_operation, which implements
OP_OBJC_MSGCALL.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* eval.c (objc_msgcall_operation::evaluate): New method.
* c-exp.h (class objc_msgcall_operation): New.

Introduce var_value_operation

This adds class var_value_operation, which implements OP_VAR_VALUE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class var_value_operation): New.
* eval.c (var_value_operation::evaluate)
(var_value_operation::evaluate_for_address)
(var_value_operation::evaluate_with_coercion)
(var_value_operation::evaluate_for_sizeof)
(var_value_operation::evaluate_for_cast): New methods.
* ax-gdb.c (var_value_operation::do_generate_ax): New method.

Implement C++ cast operations

This adds class cxx_cast_operation, which is used to implement the C++
dynamic_cast and reinterpret_cast operations.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (cxx_cast_ftype): New typedef.
(cxx_cast_operation): New template.
(dynamic_cast_operation, reinterpret_cast_operation): New
typedefs.

Introduce unop_cast_type_operation

This adds class unop_cast_type_operation, which implements
UNOP_CAST_TYPE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_cast_type_operation): New.
* ax-gdb.c (unop_cast_type_operation::do_generate_ax): New
method.

Introduce unop_cast_operation

This adds class unop_cast_operation, which implements UNOP_CAST.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_cast_operation): New.
* ax-gdb.c (unop_cast_operation::do_generate_ax): New method.

Introduce assign_modify_operation

This adds class assign_modify_operation, which implements
BINOP_ASSIGN_MODIFY.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class assign_modify_operation): New.
* eval.c (eval_binop_assign_modify): No longer static.
* ax-gdb.c (assign_modify_operation::do_generate_ax): New method.

Introduce assign_operation

This adds class assign_operation, which implements BINOP_ASSIGN.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class assign_operation): New.
* ax-gdb.c (assign_operation::do_generate_ax): New method.

Introduce type_instance_operation

This adds class type_instance_operation, which implements
TYPE_INSTANCE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class type_instance_operation): New.
* eval.c (type_instance_operation::evaluate): New method.

Introduce op_this_operation

This adds class op_this_operation, which implements OP_THIS.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class op_this_operation): New.
* ax-gdb.c (op_this_operation::do_generate_ax): New method.

Implement UNOP_MEMVAL and UNOP_MEMVAL_TYPE

This adds class unop_memval_operation and unop_memval_type_operation,
which implement UNOP_MEMVAL and UNOP_MEMVAL_TYPE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_memval_operation)
(class unop_memval_type_operation): New.
* eval.c (eval_op_memval): No longer static.
(unop_memval_operation::evaluate_for_address)
(unop_memval_type_operation::evaluate_for_address)
(unop_memval_operation::evaluate_for_sizeof)
(unop_memval_type_operation::evaluate_for_sizeof): New methods.
* ax-gdb.c (unop_memval_operation::do_generate_ax)
(unop_memval_type_operation::do_generate_ax): New methods.

Introduce unop_alignof_operation

This adds class unop_alignof_operation, which implements UNOP_ALIGNOF.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_alignof_operation): New.
* eval.c (eval_op_alignof): No longer static.

Introduce unop_sizeof_operation

This adds class unop_sizeof_operation, which implements UNOP_SIZEOF.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_sizeof_operation): New.
* ax-gdb.c (unop_sizeof_operation::do_generate_ax): New method.

Introduce unop_addr_operation

This adds class unop_addr_operation, which implements UNOP_ADDR.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_addr_operation): New.
* ax-gdb.c (gen_expr_unop) <case UNOP_ADDR>: New.

Introduce typeid_operation

This adds class typeid_operation, which implements OP_TYPEID.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class typeid_operation): New.

Introduce decltype_operation

This adds class decltype_operation, which implements OP_DECLTYPE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class decltype_operation): New.

Introduce typeof_operation

This adds class typeof_operation, which implements OP_TYPEOF.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class typeof_operation): New.

Introduce type_operation

This adds class type_operation, which implements OP_TYPE.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class type_operation): New.
* eval.c (eval_op_type): No longer static.

Introduce unop_ind_operation

This adds class unop_ind_operation, which implements UNOP_IND.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (class unop_ind_base_operation)
(class unop_ind_operation): New.
* eval.c (eval_op_ind): No longer static.  Remove "op" parameter.
(unop_ind_base_operation::evaluate_for_address)
(unop_ind_base_operation::evaluate_for_sizeof): New method.
* ax-gdb.c (gen_expr_unop) <case UNOP_IND>: New.

Implement unary increment and decrement operations

This implements the unary increment and decrement operations, "++" and
"--".

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (unop_incr_operation): New template.
(preinc_operation, predec_operation, postinc_operation)
(postdec_operation): New typedefs.
* eval.c (eval_op_preinc, eval_op_predec, eval_op_postinc)
(eval_op_postdec): No longer static.

Implement some unary operations

This introduces a couple of new template classes and then uses them to
implement some simple unary operations.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* expop.h (unary_ftype): New typedef.
(unop_operation, usual_ax_binop_operation): New templates.
(unary_plus_operation, unary_neg_operation)
(unary_complement_operation, unary_logical_not_operation): New
typedefs.
* eval.c (eval_op_plus, eval_op_neg, eval_op_complement)
(eval_op_lognot): No longer static.
* ax-gdb.c (gen_expr_unop): New function.

Introduce comma_operation

This adds class comma_operation, which implements BINOP_COMMA.

gdb/ChangeLog
2021-03-08  Tom Tromey  <tom@tromey.com>

* ax-gdb.c (comma_operation::do_generate_ax): New method.