Re: getting a copy of all characters flex has processed

[Rainer Gerhards]

Thanks for the answers! I was obviously thinking to complicated ;-) I
have now added a function to the start of processing for each and
every token. That function conditionally (depending on command line
option settings) outputs a version of everything including "include
files" to the specified file. Does the trick rather nicely, albeit at
the price of adding ~80 function calls. But that's in a very
infrequently changing area, so it doesn't really matter.

I'll probably try out YY_USER_ACTION now that the immediate need is
solved. Sounds like a more elegant solution.

Thanks again for the quick help!

Rainer

El vie., 26 abr. 2019 a las 20:22, Jannick (<address@hidden>) escribió:
>
> On Fri, 26 Apr 2019 11:26:49 +0200, Rainer Gerhards wrote:
>
> > I would like to generate that output file during regular configuration
> > processing. So in short I need a capability to see each character that
> flex
> > processes at exactly the time it processes the character. I did not find a
> > documented override to do this.
> >
> > Is there any way to handle this with flex?
>
> Add 'ECHO;' to the flex rule whose token (yytext) you want to write to
> stdout.  You might want to check the first example in
> https://westes.github.io/flex/manual/Multiple-Input-Buffers.html#Multiple-In
> put-Buffers which could be pretty close to what you currently have, but
> using flex only, not bison.
>
> Alternatively, 'gcc -E' could be a starting point which ships the nested
> include feature if the keyword 'include' was replaced by '#include'.  This
> requires the end user to have 'gcc' available, but that should not be an
> issue when dealing with 'config.h' I guess.
>
> Best,
> J.
>