Hi Gary, * Gary V. Vaughan wrote on Mon, May 22, 2006 at 01:07:19PM CEST: > Ralf Wildenhues wrote: > >* Gary V. Vaughan wrote on Mon, May 22, 2006 at 11:45:01AM CEST: > >> > >>foo=`echo X"$bar " | $SP2NL | $Xsed -e baz | $NL2SP`
> >that will still be the case, after > > | $NL2SP > > > >it will be space-space. Then, the shell is to rip off the final > >newline, if any. > > That part was deliberate on my part; I thought you were worried that > for some implementations of $Xsed input that didn't end with a NL > would be problematic... Yes. But in above sequence, that can't be the case. I think we've clarified this now. > > My point is that some shells may be confused iff > >there is no final newline. That's what the ($NL2SP; echo) would be for. > > > >I simply don't know whether that is necessary for some broken shells. > > ...but you are worried that there are shells that don't like backtick > substitutions that don't end with a newline? I would be amazed if > such a shell could even come close to being able to run a typical > configure script. I would be, too. But given how some old shells implement quoting internally (8bit stuff and the like), I wouldn't be too surprised if they were "interesting" wrt. backtick expansion, too. In fact, there is a "modern" system which has had issues in this regard: MinGW (and I assume Cygwin) have at times failed to strip the \r from a \r\n ending in such an expansion. The libtool.m4 code witnesses this at several points, somewhat inconsistently. > Or do you mean there are shells where the later > ``eval $bar'' will fail without a trailing newline? I hope not. Cheers, Ralf
