Blocker Description
On a slightly deeper look at the comparison results between 1.14 and trunk, it looks like we're missing quite a few embedded images from .doc files. I initially thought these could be explained by different handling of macros, but that's not the issue.
I haven't traced the commit that did it (very likely my fault), but...
when we call this with a null character run.
// Handle any pictures that we haven't output yet
for (Picture p = pictures.nextUnclaimed(); p != null; ) {
handlePictureCharacterRun(
null, p, pictures, xhtml
);
p = pictures.nextUnclaimed();
}
the null character run then triggers skipping of the picture in this check because isRendered(cr) returns false if cr is null
if (!isRendered(cr) || picture == null) {
// Oh dear, we've run out...
// Probably caused by multiple \u0008 images referencing
// the same real image
return;
}