This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: CPP outputs wrong number of newlines sometimes

Neil Booth wrote:
> Probably.  I'd rather not scan the string / comment again,
> particularly if we're moving towards multibyte char support.

If it's UTF-8 output, then rescanning is ok for newlines.

> It might be better to somehow record the number of lines involved when
> lexing these tokens.  That would be easy (just the difference between
> buffer->lineno before and after the relevant function call); finding
> the place to store that info would be harder.

Or just store a flag meaning "this token contains embedded newlines".
It won't be set often.  Then you can scan at output time.  I agree
recording the line number difference is cleanest, but it may not be

At the moment with -C, // comment tokens include the newline at end.  If
you don't clear pfile->need_newline in output_token, you get output like
this with redundant newlines and line directives:

  // first

# 2
  // second

However I think that can only happen with // comments, so it would be
simpler to just fix those tokens to not include the final newline

> Don't you just love GNU extensions? <g>

I don't think this is an extension.  Don't all unix preprocessors have
to do the same thing here?

> Maybe you could get the paperwork done anyway.  I'm sure it'd be
> useful for the future.

Hmm.  Which form, anyway?  So many to choose from.

-- Jamie

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]