Re: Question about token stack

[Akim Demaille]

Hi Ervin,

> Le 30 août 2020 à 11:44, Ervin Hegedüs <airween@gmail.com> a écrit :
> 
> Hi all,
> 
> and in parser:
> 
> https://github.com/airween/flextest/blob/master/myparser.y#L52
> 
> config_directive_line:
>      T_CONFIG_DIRECTIVE T_CONFIG_DIRECTIVE_ARGUMENT { printf("'%s' '%s'\n", 
> $1, $2); free($1); free($2); }
> 
> This works as well, but I found some interesting thing when I
> analyzed my code with valgrind: the pushed token is still
> reachable when the parser run into an error.
> 
> There is a wrong config: badconf.conf.
> 
> Here is the result:
> 
> valgrind -s --leak-check=full --show-leak-kinds=all --track-origins=yes 
> ./myparser badconf.conf
> 
> ==564114== 4 bytes in 1 blocks are still reachable in loss record 1 of 4
> ==564114==    at 0x483677F: malloc (in 
> /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
> ==564114==    by 0x48EB71A: strdup (strdup.c:42)
> ==564114==    by 0x10B86A: push_sym (mylexer.l:111)
> ==564114==    by 0x109E7F: yylex (mylexer.l:53)
> ==564114==    by 0x10954B: yyparse (myparser.tab.c:1055)
> ==564114==    by 0x10BD38: open_and_parse (mylexer.l:221)
> ==564114==    by 0x10BF98: main (mylexer.l:272)
> 
> 
> How can I avoid this situation? Is that any "best-practice"? How
> can I access to pointers in stack to free them when error
> occurred?

I have not read all in details, but I believe you need to read
about %destructor.

https://www.gnu.org/software/bison/manual/html_node/Destructor-Decl.html

Cheers!