On Wed, Jul 14, 2010 at 02:44:23PM +0200, Daniel Näslund wrote:
> Hi!
>
> I just realized that parse_next_hunk() at present only detects a property
> hunk if it has a property header ('Added', 'Deleted', 'Modified')
> directly preceeding it. The second hunk won't know what propname it
> belongs to.
>
> We previously had diff headers and hunk headers. The property header is
> only set before the first hunk and is thus not the same sort as any of
> the two previous ones.
>
> What should parse_next_hunk() do to properly handle property hunks?
> ---------------------------------------------------------------------
> Alternative 1
> Fetch the next hunk and tell the caller if it's a property or a text
> hunk. (let it be up to the caller to remember what the last prop_name
> was)
I think 1 sounds better. It seems better to let parse_next_hunk() parse
hunks, and let the caller worry about additional context.
>
> Alternative 2
> Keep track of if we're currently below a property header (and hasn't
> seen any text hunks since that property header). That means passing
> around a 'prop_name' variable that would be set to NULL if the previous hunk
> was a text hunk and otherwise have the value of the last parsed
> property name.
>
> Any suggestions?
>
> Thanks,
> Daniel
