Re: Bug report . . . maybe

[Hans-Bernhard Broeker]

On Mon, 17 Sep 2001, Mike Whiteley wrote:

> I noticed that if a lex input file was edited on DOS or windows.
> (causing there to be a <CR><LF> pair) this causes character class
> definitions to no longer work on my Linux box.

Well, the obvious answer would then be "don't do that, then", right?  
It's an error to assume you can use textfiles interchangeably between
DOS/Windows and Unix without having to convert them on transfer.  The guys
who put the 'ascii' feature into the ftp protocol knew what they were
doing, see? 

And this is by no means limited to (f)lex. The same kind of problem used
to happen with plain C files, too (backslash-newlines break if the newline
becomes a CRLF). GCC only learned about this around version 2.95, IIRC.

Of course, flex could be changed to be more picky about non-printable
characters, but do we really want that. E.g., should/can we forbid literal
ASCII control characters (chars < 32) in lex input? And what about
non-ASCII charsets, while at that?

-- 
Hans-Bernhard Broeker (address@hidden)
Even if all the snow were burnt, ashes would remain.