#10 hyperlinks misparsed

[Andrew Pimlott]

I found some issues with parsing hyperlinks. Actually, I found them in an
older, hacked version of majix. But they seem to apply to the current
version of makix. However, I haven't inspected the current version as
closely, so forgive me if I get some things wrong. Also, I don't really
know anything about RTF--this is based on debugging problems in one file.

The code seems to expect hyperlinks to look something like this:

{\field {\*\fldinst ... HYPERLINK ...}
{\fldrslt ...}}

But in the file I have they look like:

{\field {\cf15 {\*\fldinst ...}}
{\cs5 {\fldrslt ...}}}

I don't know whether this is legal, but I have to handle it. (The file says
{\*\generator Microsoft Word 11.0.0000;}).

Let's see how this throws things off: In processField, we pass by {\cf15
(incrementing depth), and when we find \fldinst we call processFieldInstance
to do the rest. Here we've already lost, because processFieldInstance is
expected to read through the close of \fldinst and into \fldrslt; but it
doesn't know it has to go up a level past the \fldinst and close the \cf15.
What happens is getDataOfGroup reads through the close of \fldinst, and when
we get to getResultOfField, the first token we read is a } and we return
null.

In some cases, they look like:

{\field {\*\fldinst ...}
{\cs5 {\fldrslt ...}}}

In this case, we start getResultOfField at {\cs5 and skip past to {\fldrslt.
But then we return, without reading the close of \cs5 (because
getResultOfField does not track depth). That close will eventually be read
by parse, which will pop _theTextPropertiesStack. Note that
getResultOfField was not pushing, so that's a pop without a push.

I have a patch against the old version of majix I am using. I can supply a
patch for the current version, if you're interested. But since I might be
off base, I want to ask for guidance first.

Andrew

[Nobody/Anonymous]

http://unseddaweb.hotbox.ru