Re: [pdf-devel] [PATCH] token reader unit test update

[Michael Gold]

On Tue, Nov 03, 2009 at 22:02:21 +0100, address@hidden wrote:
> 
>    I don't think it would be useful to reject a file for violating
>    this rule at EOF; but if we decide to do that, we should do it in a
>    more general way since the rule isn't specific to any token type.
> 
> Would be possible to, while permitting to violate the rule at EOF,
> make the tokeniser aware of the problem and raise some flag?  In that
> way we could warn the user about the absence of an EOL before EOF.

It wouldn't be difficult to record the last character seen.  If it's not
an EOL character, and the token reader is in the EOF state, the file
contains this error.

I don't see what users would do with this information.  Is a missing EOL
at EOF known to cause problems with any software?  It's not clear to me
that the PDF specs intended to define it as an error (e.g., PDF32000
7.5.1 describes the usage of lines as "a matter of convention"; and
since this section doesn't apply to content streams, they don't need EOL
markers).

-- Michael

signature.asc
Description: Digital signature