Simon Wright <[email protected]> writes:
> If I put my cursor on line 101 ant press TAB, the message shown by C-c
> C-f is
point location when you invoke C-c C-f is irrelevant; point is moved to
the error.
> normalize_xmi-model-association_classes.adb:116:37: too many
> parallel parsers required; simplify grammar, or increase
> wisi-parse-max-parallel’
I can reproduce this.
> wisi-parse-max-parallel is set to 15, I tried setting it to 30 but no
> change.
That moves the error deeper in the code; going to 60 allows the parse to
succeed (with a noticeable delay).
> I realise that this code is one where insisting on max line length of
> 79 makes for uncomfortable reading. Any thoughts on how it might be
> reorganised?
Subroutines for each case, instead of a multi-nested case statement.
Replacing the leaves in 'case E1.Lower' with subroutines fixes the
parsing.
Using ada-indent-when = 0 would also help readability.
A better fix is to remove the grammar ambiguity. It is due to allowing
an empty list of alternatives in a case statement:
case foo is
end case;
The intent of this is to make it easy to type a new case statement that
passes the parser.
On the other hand, the 'case' skeleton (in ada-skel.el) expands to:
case is
when =>
end case;
which has a 'when', so it doesn't require this relaxation.
However, the skeleton expansion does not have an expression after
'case', so it fails the parser. That's easily fixed by changing
'expression' to 'expression_opt' in the case_statement in
ada-grammar.wy.
My stated requirement on the relaxed syntax in the parser is to succeed
on all code generated by ada-skel. Apparently I need to write more tests.
I think it's clear I should drop allowing a case with no 'when', in
order to support deeply nested case statements.
Here's a patch (also in ada-france):
============================================================
--- ada-grammar.wy 2b4ac85eae3684001a5dbecc8b1bb44f1e59d5cc
+++ ada-grammar.wy 4115b2ce1598091c40692cf4971dffa1a569c774
@@ -208,7 +208,6 @@
%conflict SHIFT/REDUCE in state actual_parameter_part, attribute_designator on
token LEFT_PAREN
%conflict SHIFT/REDUCE in state association_opt, name on token EQUAL_GREATER
%conflict SHIFT/REDUCE in state attribute_reference, attribute_designator on
token TICK
-%conflict SHIFT/REDUCE in state case_statement_alternative,
case_statement_alternative_list on token WHEN
%conflict SHIFT/REDUCE in state formal_derived_type_definition,
formal_derived_type_definition on token WITH
%conflict SHIFT/REDUCE in state label_opt, label_opt on token IDENTIFIER
%conflict SHIFT/REDUCE in state name, discrete_subtype_definition on token
RIGHT_PAREN
@@ -476,7 +475,7 @@ case_statement
;
case_statement
- : CASE expression IS case_statement_alternative_list END CASE SEMICOLON
+ : CASE expression_opt IS case_statement_alternative_list END CASE SEMICOLON
(progn
(wisi-statement-action 1 'block-start 3 'block-middle 5 'block-end 7
'statement-end)
(wisi-containing-action 1 2)
@@ -492,9 +491,11 @@ case_statement_alternative
(wisi-containing-action 3 4))
;
+;; We don't allow an empty list here; that leads to parallel parser
+;; explosion in nested case statements. Note that ada-skel-case
+;; inserts an empty 'when =>'
case_statement_alternative_list
- : ;; empty
- | case_statement_alternative
+ : case_statement_alternative
| case_statement_alternative_list case_statement_alternative
;
--
-- Stephe
_______________________________________________
Emacs-ada-mode mailing list
[email protected]
http://host114.hostmonster.com/mailman/listinfo/emacs-ada-mode_stephe-leake.org
