Question

I created test.l, input to flex, which ends with the main function.

When the main function is implemented as:

int main(void) {     yylex();     return 0; } 

I have no problem.

I want to trick the parser into believing that the first character is always a semi-colon, so I implemented main as

int main(void) {     unput(';');     yylex();     return 0; } 

the above leads to a segment fault.

Why does the use of unput lead to a segment fault?

Answer 1

It is causing a segment fault because yylex() has not yet initialized the input buffers etc. that are needed by unput().

There’s probably a better way to design your scanner without needing to trick it into a leading semicolon, but if you must do this then one solution might be to use start conditions. Something like this:

%x SPECIAL %%     BEGIN(SPECIAL); /* Go to SPECIAL state when yylex first called */  <SPECIAL>.  { unput(*yytext); unput(';'); BEGIN(INITIAL); }  ... rest of rules ...