Re: Parsing '\x00' -- appears to be the only thing preventing parsing binary data

[Stephen Taylor]

Thanks, that seems to get around \x00 -- unfortunately, the same issue then 
reappears with the values \x80 - \xff … I could special case every single value 
e.g.

\x00 {return ZERO; }
\x80 { return X80; }
\x81 { return X81; }
 :
\xff { return XFF; }
. | \n  { return *yytext; } 


But seems a bit clumsy. Any thoughts?


> On Mar 30, 2021, at 17:01, Chris verBurg <cheetomonster@gmail.com> wrote:
> 
> 
> Running bison (first) with -d should produce a bison.tab.h file, that your 
> flex input should #include to get the definitions of %token names.
> 
> -Chris
> 
> 
> 
>> On Mar 30, 2021, at 1:34 PM, Stephen Taylor <stnh.email@icloud.com> wrote:
>> 
>> 
>> That appears to give me an error in lex:
>> 
>> %option noyywrap                                                             
>>                                                      
>> 
>> %%                                                                           
>>                                                      
>> \x00  { return ZERO; }                                                       
>>                                                      
>> .|\n  { return *yytext; }                                                    
>>                                                      
>> %%                                                                           
>>                                                      
>> 
>> 
>> —
>> Error:
>> 
>> flex -8 gmr.l
>> gcc  -c lex.yy.c
>> gmr.l: In function ‘yylex’:
>> gmr.l:2:10: error: ‘ZERO’ undeclared (first use in this function); did you 
>> mean ‘EIO’?
>> \x00  { return ZERO; }
>>          ^~~~
>>          EIO
>> gmr.l:2:10: note: each undeclared identifier is reported only once for each 
>> function it appears in
>> Makefile:13: recipe for target 'lex.yy.o' failed
>> make: *** [lex.yy.o] Error 1
>> 
>> 
>> Bison likes it.
>> Tried quoting the \x00 but thats not right it appears.
>> 
>> 
>> 
>>> On Mar 30, 2021, at 16:15, Chris verBurg <cheetomonster@gmail.com> wrote:
>>> 
>>> 
>>> I don’t know how you’ve set up your flex/bison interface, but I’m guessing 
>>> from your quoting that you have flex return each character as a token?  Can 
>>> you special-case 0 to not have the token value 0?  That is, flex does this:
>>> 
>>> \x00  { return ZERO; }
>>> 
>>> And then in your grammar:
>>> 
>>> %token ZERO
>>> 
>>> and use ZERO instead of ‘\x00’?
>>> 
>>> (Note: typed this on a phone..not 100% sure the syntax is right.)
>>> 
>>> -Chris
>>> 
>>> 
>>>> On Mar 30, 2021, at 12:37 PM, Stephen Taylor via Users list for the GNU 
>>>> Bison parser generator <help-bison@gnu.org> wrote:
>>>> 
>>>> I am trying to parse binary files and obviously need to write rules that 
>>>> include the 8-bit value 0x00 — flex appears to allow me to obtain an 8-bit 
>>>> scanner using the -8 option. I seem to be able to use ‘\x01’ to ‘\xFF’ for 
>>>> individual terminal symbols in Bison rules. Unfortunately, ‘\x00’ is used 
>>>> to denote end of input. Looking at the output tables, there appears to be 
>>>> no real reason to do this (other than convenience) and I am wondering if 
>>>> anyone has a workaround? 
>>>> Thanks !
>>>> Steve
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>

smime.p7s
Description: S/MIME cryptographic signature