On Thu, Oct 17, 2013 at 1:58 PM, Liam Healy <l...@healy.washington.dc.us> wrote: > I'm not even sure why e-t-f-d-indirect doesn't always just call > e-t-f-d except where the type is directly represented, but I'm not > going to mess with something that works.
I guess it's because of that exception since we have to deal with user-defined types. Perhaps if the types could be characterised as immediate versus pointer/aggregate (or something like that) then we could be smarter about e-t-f-d-indirect. Anyway, a test using libtest (or some standard C function, if we can force libffi to kick in for functions that don't use structures) would be nice. Do you have the time/inclination to give it a go? -- Luís Oliveira http://kerno.org/~luis/
