On Thu, Feb 10, 2005 at 10:21:19PM -0500, Ian Lance Taylor wrote: > I found other cases where an empty vector produces a NULL pointer and > leads to a later crash. So I don't think it is safe to have an empty > vector in RTL for the 'E' format.
This surprises me not at all. FYI, typically we put (const_int 0) into the vector when we've nothing else meaningful. > I propose this patch. OK if it passes testsuite and bootstrap? Yes. r~
