On Mon, May 12, 2025 at 11:42:43PM +0200, Ingo Schwarze wrote:
> Theo Buehler wrote on Mon, May 12, 2025 at 10:03:50PM +0200:
> > (If someone copies this chunk of code, they might not notice this
> > assumption
> 
> Someone copying this code seems unreasonable to me, given how
> specialized it is: "strip any trailing slashes and the last filename
> component from the end, then iterate the ends of all other
> components" really doesn't sound like something that might be
> needed outside "rmdir -p".
> Then again, an assertion does prevent overlooking this invariant.
> 
> > and a check in code is always better than a comment.)
> 
> I disagree.  Both serve more or less orthogonal purposes.
> Checks guard against functional edge cases.  Comments help
> code review and maintenance.

It would also help if we ensure that the CVS commit message makes it clear
that the issue was discussed at length on-list.

Since our proposed implementation differs from the other BSDs, there is a at
least a reasonable chance that anybody interested in the code will look
through CVS to see when and why the changes were made.