Redo the whole thing again.

Parsing the chunks separately was not a good idea. The file layout seems
to be alternating ROOT and ARCH chunks, where the ROOT chunk describes
what kind of data to expect in the following ARCH chunk.

That is:

    +------+------+------+
    | RIFF | SIZE | NUND |
    +------+------+------+------+ \
    | ROOT | SIZE | NAME | TYPE | |
    +------+------+------+------+  > One unit
    | ARCH | SIZE | DATA....    | |
    +------+------+-------------+ /
              :
              :
    +------+------+------+------+ \
    | ROOT | SIZE | NAME | TYPE | |
    +------+------+------+------+  > Last unit
    | ARCH | SIZE | DATA....    | |
    +------+------+-------------+ /
    EOF

The NAME and TYPE are counted (and possibly zero terminated) strings
describing the name of the following chunk (e.g. "Version"), and data
type (e.g. "PAppVersion") respectively.

The order of the chunk pairs does not seem significant, but the pairs
are always a ROOT chunk followed by an ARCH chunk.

0 files changed, 0 insertions, 0 deletions