IMO, there is little to no reason to complexify the code and add to the users' cognitive burden by adding a new option. That written, NO_CLOBBER is a feature I don't use, so I'll contentedly end my part of this conversation now. - Vin On Wed, Jun 3, 2020 at 10:36 PM Daniel Shahaf wrote: > Vin Shelton wrote on Wed, 03 Jun 2020 22:13 -0400: > > What is the significance of a 0-length file as opposed to a file with > > contents? > > There _are_ some cases where a 0-length file is treated differently to > a non-existing one: > > - .vimrc > - «cat foo» (also #include's in C) > - GNU make(1)'s lookup order: GNUmakefile / Makefile / makefile > - /etc/nologin > > > If we want to support this feature, why don't we change the meaning of > > NO_CLOBBER to mean only non-empty files? > > Because users might be relying on the semantics that the documentation > promises. > > For example, a user who does «echo foo > /etc/nologin» and doesn't get > an error may infer that it's fine to delete the file once they finish > whatever change they're working on. Keeping the error would alert that > user that they shouldn't remove the file when they're done (and, most > likely, should pause to coordinate with whoever created /etc/nologin > before them). > > Cheers, > > Daniel > > > > - Vin > > > > On Wed, Jun 3, 2020 at 8:05 AM Peter Stephenson < > p.stephenson@samsung.com> > > wrote: > > > > > Martin Tournoij wrote: > > > > I switched from tcsh to zsh a while ago (many years too late, I > know), > > > > and found zsh can do pretty much everything better. There's one thing > > > > I rather miss though: the 'notempty' option in 'noclobber'. > > > > > > This isn't actually hard to implement. What does everyone else think? > > > > > > pws > > > > >