Changes made to ensure that '=>' token is parsed properly #3292
Conversation
There was a problem hiding this comment.
Thanks @sriganeshres for contributing!! Could you add a testcase to your PR to ensure we don't introduce a regression on this in the future?
sriganeshres
force-pushed
the
chore-parse-expr
branch
2 times, most recently
from
b86ad1a to
62abc58
Compare
Working on that. I have ran Thank You. |
sriganeshres
force-pushed
the
chore-parse-expr
branch
7 times, most recently
from
5aca0e5 to
7897d6e
Compare
gcc/rust/ChangeLog: * parse/rust-parse-impl.h (Parser::parse_expr): Updated parsing logic to stop on encountering '=>' or ',' while having restrictions.stop_on_token turned on. (Parser::left_denotations): Updated to stop parsing on the special tokens mentioned above. (Parser::null_denotation): Updated to stop parsing on the special tokens mentioned above. * parse/rust-parse.h (struct ParseRestrictions): Added a field stop_on_token (default false) to maintain the above functionality.
sriganeshres
force-pushed
the
chore-parse-expr
branch
from
7897d6e to
e7e7360
Compare
You could write a test from the snippet given in the fixed issue, it would pass with your changes |
* rust/compile/issue-3099.rs: Added a new test for issue-3099 to parse '=>' properly.
| @@ -89,6 +89,8 @@ struct ParseRestrictions | |||
| bool entered_from_unary = false; | |||
| bool expr_can_be_null = false; | |||
| bool expr_can_be_stmt = false; | |||
| bool stop_on_token | |||
| = false; // This is for stopping parsing at a specific token | |||
There was a problem hiding this comment.
What does this mean? It seems like a hack
| _ => 20, | ||
| }; | ||
|
|
||
| let format = "Result: %d\n\0" as *const str as *const i8; |
There was a problem hiding this comment.
hi @sriganeshres, make sure that you also update your dejagnu regex to assert that you'll get out 15 from println!.
Without asserting the printf, you're just guarantee that we at least parse it successfully. Or successfully breaking something in the parser silently
| } | ||
|
|
||
| pub fn main() { | ||
| unsafe { |
There was a problem hiding this comment.
is the unsafe necessary?
There was a problem hiding this comment.
it's necessary for the printf call, but not before - so it could be moved down a few lines indeed
| @@ -12023,6 +12023,11 @@ Parser<ManagedTokenSource>::parse_expr (int right_binding_power, | |||
| if (restrictions.expr_can_be_null) | |||
| { | |||
| TokenId id = current_token->get_id (); | |||
| if (restrictions.stop_on_token && (id == MATCH_ARROW || id == COMMA)) | |||
There was a problem hiding this comment.
you're not setting stop_on_token to true so you're not entering these conditional either.
| if (restrictions.stop_on_token | ||
| && (current_token->get_id () == MATCH_ARROW | ||
| || current_token->get_id () == COMMA)) |
There was a problem hiding this comment.
this seems weird. you'd want to enter left denotation and create the infix with =>
gcc/rust/ChangeLog:
Fixes #3099
Here is a checklist to help you with your PR.
make check-rustpasses locallyclang-formatgcc/testsuite/rust/Parser::parse_expr,Parser::left_denotations,Parser::null_denotationSo that they stop on special token like=>and,. They would terminate the parsing to allow exatract expression.