On Fri, Oct 08, 2021 at 10:21:26PM +0200, Stefan Sperling wrote:
> Another way of "fixing" the issue without a major rewrite is to read
> data from input files in units of lines rather than BUFSIZ chunks.
> We can then prepend a symbol of our choice (other than a dot) before
> each line which represents content that is supposed to appear in the
> merged output. Now all content is "escaped" and will never result in a
> dot on a line by itself that can be mistaken as a delimiter. We strip
> our leading symbol before writing a line of content out.

Does anybody have a better idea, or should I go ahead with this hack?