Lists are really crazy. Lists information is stored seperately from the entries
involved, and indexed in the pap. I have come across negative values for this!,
normalizing them to postive appears to do the right thing. Also I have reworked
the "invalid list entry" code a bit, this is where the format promises that a
number of list overrides are forthcoming in the file, and then inserts
0xffffffff as some sort of marker. Maybe it means that "there are no list
overrides after all", but I believe that it just means "ignore me and try
again after me". Following that rule creates the possibility of heading off
the end of the file, so a test for that condition is also included. This wipes
out a few more broken files.
Tables have a table look specifier which sets what way the table should be
colored in, and what sort of borders it should have. There are 40 table
looks that can be selected. Needless to say there are files that index a
value way above 40, this also has to be caught and munged into a valid
number.
Both the ilfo and tlp sanity checks will be required by a program using
the wv library. The work arounds are not part of the core library in case
I have them completely wrong. So when abiword starts using lists and table
looks they will have to sanity check the ilfo and tlp values. Basically I
reccommend getting back to me when this functionality is being put into
abi's half of the word converter.
Also I have added a few sanity checks here and there, primarily just so
that I could get one file to convert that was crashing MSWord97 itself. So
that I could feel smug about it all :-)
This basically means that out of approximitately 2500 word 8 files that wv
does not crash at all. Though your milage will vary as this is defined by
not triggering a SIGSEGV on my machine. There might well be bugs that will
still crash on more intolerent systems. And of course there is no guarantee
that the output files are sane. There could be loads of madness there yet.
And I havn't run the tests over word 6 and 7 files yet. Still I'm considering
this a milestone. More and more the crashes are identifying wierd word
behaviour rather than weird programmer behaviour.
Another added program is wvVersion which reports the version of the word
document given to it as an argument. It will be possible in the future to
support word documents prior to the ole2 change over. i.e. word 5 through 1,
Im putting down some of the framework to support this at the moment.
Current work under way is
1) stylesheet ordering
2) more word 6 and 7 support
3) graphics
4) automatic html validation testing as part of the crash testing for wvHtml,
this should confirm that the output of wvHtml is sane.
Once there three are completed to my satisfaction I'll have a stable release
and remove all traces of mswordview in favour of wvHtml. Which will be nice.
Future plans include
support for older word version
some save as word support
move the ole2 code over to the libole2 library that some of the gnumeric and
cole people are working on at the moment. Should mean a speed up when this
gets folded in.
Latest versions are in abi's cvs and version 0.5.38 at
http://www.csn.ul.ie/~caolan/publink/mswordview/development/
C.
Real Life: Caolan McNamara * Doing: MSc in HCI
Work: Caolan.McNamara@ul.ie * Phone: +353-86-8790257
URL: http://www.csn.ul.ie/~caolan * Sig: an oblique strategy
Mechanicalise something idiosyncratric