Chromatic <[EMAIL PROTECTED]> wrote: > That is, to make the keyed struct work correctly, I had to add extra > bytes of padding in the appropriate places.
Have a look at the third initializer param - this is the offset of the item in bytes. (Albeit untested - seems you got the code to test it :) > Is this indeed unportable? If so, is there a way to work around this > without writing C code for every afflicted structure? Is it a case of > making the struct PMCs smarter? Or have I fixed it with my nasty > workaround? NCI is per se unportable. But the *Struct PMC could of course be smarter and calculate common align requirements. There are of course still issues if you have to deal with packed structures. > -- c leo
