X && Y || Z
binds operator precedence-wise as
(X && Y) || Z
not
X && (Y || Z)

so I am a little suspicious about...
source/dialog/numpages.cxx
where we have repeated patterns of...

if(bPreset || bModified && pActNum)
    *pSaveNum = *pActNum;

i.e. its clearly unsafe to dereference pActNum if it is NULL if bPreset if true
and bModified is false, but that is what would happen at the moment.

Are we sure that's what we want, if not the attached patch fixes them up.

Also note the first instance tests off pSaveNum which is different to the normal
pattern of testing pActNum, but otherwise the code is the same, which is also
suspicious