Hrvoje Niksic <hniksic(a)srce.hr> writes:
Not if the lvalue is supposed to be read-only. I prefer a crash to
silent problems.
Then mprotect the purespace on ELF systems.
If we agree that we don't write to purespace, then we should not
use
XC[AD]R as lvalues, regardless of whether the coredump will happen if
we slip.
Or to simply not put cons cells in the purespace. It IS only 80K.
What function call? Have you read my proposal on how to transform
the
code that uses XCAR/XCDR as lvalues? (hint: I don't propose reverting
to Fsetcar().)
Yes. However I think that going into all that fault handling stuff is
not worth it just to add 80K in shared pages. Page faulting would be
nice to have for an incremental garbage collector though.
Jan