On 01/04/13 18:25, Eric Blake wrote: > On 01/04/2013 05:47 PM, Bruce Korb wrote: >> You may have been thinking you were using "memcpy", but you were >> using "bcopy" instead. Please apply the patch to md5.c. Thanks! > > Rather than fixing the call site of memcpy, we should instead fix the > real bug, the fact that you are even doing *ifndef HAVE_MEMCPY and > defining memcpy to bcopy() in the first place.
I am removing (have removed) md5.[ch] from the sharutils repo. But I have a theory now: Long ago and far away, there used to be a test for memcpy. Somewhere along the line, HAVE_MEMCPY disappeared and sharutils kept working because bcopy works just fine. Then I upgraded to gnulib's md5 and I hit a booby trap and my $'^[ \t]*#[ \t]*define[ \t]+memcpy\>' re didn't find the define. It is still ugly to nest functional functions into a single line of code. But I don't care any more. The booby trap lived in sharutils/lib/system.h along side a *LOT* of ancient cruft.
