I'm using %glr-parser and ran into a snag with the %merge feature.
I figure I must be misinterpreting something along the way.
I think I've done what the docs say:
https://www.gnu.org/software/bison/manual/html_node/Merging-GLR-Parses.html
Everything compiles fine except one bison generated line in the parser code.
Assuming I have something simple like the following:
%token <some_type> IDENTIFIER
%type <some_type> subrule expr
...
subrule:
IDENTIFIER { $$ = $1; } %merge <stmtMerge>
| expr { $$ = $1; } %merge <stmtMerge>
;
The Issue:
Bison is generating the following 'yyuserMerge' code (near the bottom of this
message).
Look at the case statement in the procedure.
it generates:
case 1: yy0->some_type = stmtMerge (*yy0, *yy1); break; // Assigning to a
type?
^^^^^^^^^^^^^^ Here
Instead of something like:
case 1: *yy0 = stmtMerge (*yy0, *yy1); break; // This *might* make sense. Looks
better anyways :)
^^^^ Here
What is actually happening does not make sense to me.
'some-type' is treated as a variable name. Its a type.
The subrule has a type. It is not a variable.
Considering YYSTYPE is the 'semantic_type', it also does not make
sense to have this 'semantic_type' as a field inside itself.
I'm confused. Not sure where to look for a solution.
This is the resultant code generated by bison:
static void
yyuserMerge (int yyn, YYSTYPE* yy0, YYSTYPE* yy1)
{
YYUSE (yy0);
YYUSE (yy1);
switch (yyn)
{
case 1: yy0->gen::index_t = stmtMerge (*yy0, *yy1); break; // My only compile
error :(
default: break;
}
}
where: gen::index_t is the 'some_type' for the rule I used.
The %merge is in that rule.
Ok, now feeling awkward / foolish / dumb ...
Can anyone tell me what I am doing wrong?
This is my last issue that I see to make my project compile.
Thanks for any help / guidance anyone can give.
