sexp_make_exception can clobber SEXP_G_OOM_ERROR

[dcolascione]

sexp sexp_make_exception (sexp ctx, sexp kind, sexp message, sexp irritants,
                          sexp procedure, sexp source) {
  sexp exn = sexp_alloc_type(ctx, exception, SEXP_EXCEPTION);
  sexp_exception_kind(exn) = kind;
  sexp_exception_message(exn) = message;
  sexp_exception_irritants(exn) = irritants;
  sexp_exception_procedure(exn) = procedure;
  sexp_exception_source(exn) = source;
  sexp_exception_stack_trace(exn) = SEXP_FALSE;
  return exn;
}

What happens if sexp_alloc_type returns SEXP_G_OOM_ERROR? Now you're mutating the singleton before you return it.

[dcolascione]

Wait a second. Tons of things in this codebase call alloc functions without checking whether they're returning exceptions. Looks like memory exhaustion can cause generalized memory corruption here. Shouldn't these places be fixed

[ashinn]

Yes, the OOM error was an afterthought and is not consistently checked for. Checking for it everywhere could be a huge undertaking and inflate the size of the source code and executable, though, I need to audit it.

[dcolascione]

Sure. Abort-on-OOM is probably workable too; it's just that the partial support and UB if you're unlucky seems like the worst of both worlds.