Closed curldivergence closed 1 year ago
Looks like it is some weirdness in exception handling, most likely by doctest
library (everything is happening inside a doctest
's TEST_CASE
): apparently I had some uncaught std::bad_any_cast
s in my semantic actions, and instead of reporting those, both console output and debugger for some reason reported this assertion inside cpp-peglib
. After I started catching the bad any_cast
s locally and using a regular assert()
to break the program right at the place where the bad cast happened (as opposed to doctest
catching it and reporting in console), the assertion inside cpp-peglib
stopped reproducing.
So I think that cpp-peglib
is not to blame - it was just being taken by surprise unwinding in the wrong moment :)
For now I'll live with this assert
workaround, and if I have time to get to the bottom of this, I'll report it to doctest
people.
Sorry for bothering you, I'm closing the issue :)
No problem. Thanks for letting me know the story. Hope your project goes well!
Hi! First of all, thanks for the awesome library, I really enjoy using it. However, I've run into a problem: at some point the parsing started to fail with the assertion mentioned in the title. So I'm curious: does this assertion indicate a problem with the user-provided grammar, or something goes wrong inside the library? And if the issue lies in the grammar, are there any bits of advice you could share with me that could help debugging it? Also, I've tried parsing the same code on the online playground, and it didn't fail (but I'm not sure if the version there has assertions enabled).
For the reference,
here is my grammar
``` Program <- Item* Item <- Function / StructDeclaration StructDeclaration <- 'struct' Identifier '{' MemberDeclaration* '}' MemberDeclaration <- Type Identifier ';' Type <- ReferenceType / ValueType ValueType <- Identifier ( '<' TypeList '>' )? ReferenceType <- ValueType '&' TypeList <- Type (',' Type )* Statement <- Initialization / Assignment / Return / ExpressionStatement / Block Initialization <- Type Identifier '=' Expression ';' Assignment <- Expression '=' Expression ';' Return <- 'return' Expression ';' ExpressionStatement <- Expression ';' Expression <- InfixExpression(PrefixExpression, InfixOperators) InfixExpression(AtomicExpression, Operator) <- AtomicExpression (Operator AtomicExpression)* { precedence L == != L < > L + - L * / } If <- 'if' '(' Expression ')' Block ( 'else' Block )? Function <- Type Identifier '(' Parameters ')' Block Parameters <- List(Type Identifier, ',') Block <- '{' ( Statement )* '}' Call <- Primary (Arguments / Index / MemberAccess)* Arguments <- '(' List(Expression, ',') ')' Index <- '[' Expression ']' MemberAccess <- '.' Expression PrefixExpression <- PrefixOperator* Call Primary <- Identifier / If / Function / Array / Hash / Decimal / Integer / Boolean / String / '(' Expression ')' Array <- '[' List(Expression, ',') ']' Hash <- '{' List(HashPair, ',') '}' HashPair <- Expression ':' Expression Identifier <- < !Keyword [a-zA-Z][a-zA-Z0-9_]* > Integer <- < [0-9]+ > Decimal <- < [0-9]+ '.' [0-9]* > String <- < ["] < (!["] .)* > ["] > Boolean <- 'true' / 'false' PrefixOperator <- < [-!] > InfixOperators <- < [-+/*<>] / '==' / '!=' > Keyword <- 'true' | 'false' | 'return' | 'if' | 'else' List(Item, Delimiter) <- (Item (~Delimiter Item)*)? LineComment <- '//' (!LineEnd .)* &LineEnd LineEnd <- '\r\n' / '\r' / '\n' / !. %whitespace <- ([ \t\r\n]+ / LineComment)* %word <- [a-zA-Z]+ ```(it's a bit vandalized to fit my purposes grammar from your implementation of the
monkey
language).The code that leads to the assertion is fairly trivial:
I think that the issue is somehow related to parsing the
Expression
rule, since if line2+3;
is deleted, the assertion goes away, too, but so far I didn't manage to pinpoint the exact issue.Thank you!
UPD: Immediately after posting this I tried removing all my semantic actions (sorry, did not think of that earlier), and the assertion went away, too, so it's definitely me doing something wrong :) However I still have no idea what I'm doing wrong, so if you have any pieces of advice, I'll very appreciate if you shared them.