Hi there, Am Dienstag, den 23.05.2006, 16:47 +0200 schrieb Robert Schuster: > I make this to a general discussion. > > >> Agreed, and I think we should replace existing NSA code with the > >> simpler long field approach. Ideally we could remove the NSA > >> implementation altogether. I'd also like to see > >> Pointer/Pointer32/Pointer64 gone but they seem to be used in places > >> other than the peers. > Does that mean we simply store any native pointer as long in some field of a class? > > If so, do we use JNI to set/get the fields or do we go the simpler way provide > necessary pointers via argument list? > > I vote for the second approach because I dislike nothing but these tedious JNI > findClass, get fieldID and set field function stuff. > My preferred approach is to store pointers as Object (and access some field via index), which has the advantage that you can't fiddle with the pointer from the Java-side in any way. Of course, using a long is slightly more clear. For the access of the pointer I prefer to pass it directly via the method arguments. This avoids all the find* blah stuff. If this ain't possible for some reason it is advisable to only do the findClass stuff once and store the indices in some variable (IMO). /Roman -- ?Improvement makes straight roads, but the crooked roads, without Improvement, are roads of Genius.? - William Blake -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 191 bytes Desc: Dies ist ein digital signierter Nachrichtenteil Url : http://developer.classpath.org/pipermail/classpath/attachments/20060524/8235c753/attachment.pgp
