13817 - xpconnect assumes that getter preceeds setter

Status: VERIFIED FIXED

(Core :: XPConnect, defect, P3)

Description

[John Bandhauer]

When typelibs have attributes that are not readonly (i.e. have both a getter and
setter) then xpconnect ASSUMES that the getter will come first followed
immediately by the setter. The reason xpconnect imposes this restriction has to
do with the fact that we don't support writeonly. xpconnect walks through the
methods building its own representation. It uses one struct for attributes and
marks them as having a getter and/or setter. I didn't want to have a state for
these structs where there is a setter but no getter. While this is not a valid
'final' state, if the setter were to preceed the getter in the typelib then the
struct would 'pass through' that state as this information is read from the
interface info manager.

The 'immediately' preceeding part is so that as building this array of info
xpconnect does not have to do a lookup to match a getter with a setter.

I could work around all this at some cost or leave it as is with this
restriction.

Is this completely whacked? Should we specify this restriction in the typelib
format or 'fix' xpconnect? I'd rather specify this in the typelib and be done
with it, but I'm open to suggestion.

Comment 1

[John Bandhauer]

code is at
http://lxr.mozilla.org/seamonkey/source/js/src/xpconnect/src/xpcwrappednativecla
ss.cpp#158

adding fur so as to waste some of his time and perhaps get an opinion :)

Comment 2

[John Bandhauer]

Well, I'm thinking that it was a waste of time to even write this all up.
Considering the goals of the typelib spec - to support xpconnect - I'm just
going to add this as a requirement in the spec and be done with it!

Comment 3

[John Bandhauer]

Modified spec.

Comment 4

[cbegle]

i see you're taking bug fixing lessons from fur!