Eric Blake <[EMAIL PROTECTED]> wrote: ... > | +dnl See if CR is the EOL marker. If not, remove any EOL-related > | +dnl ^M bytes and escape any remaining ones. If so, just use mv. > > If CR is _not_ the EOL marker, then how would CR appear at the end of the > line? Where are we doing input into $tmp/subs1.awk that could create a > spurious CR if the platform does not use CR? And if CR is not EOL, then > do we really need to escape intermediate CR rather than treating it as a > literal character?
Before writing the patch, I wondered about that too, and did a little archaeology, but didn't find any good rationale, though I didn't look very hard. > And according to this patch, if CR _is_ the EOL marker, then just using mv > works. So why not use mv unconditionally? Assuming the existing CR-conversion code (not just for EOL markers) was necessary, I thought it best to minimize changes for the common EOL==NL case.
