On Wed, Sep 23, 2009 at 06:08:46PM +0200, Corinna Vinschen wrote: >On Sep 23 10:09, Christopher Faylor wrote: >>On Wed, Sep 23, 2009 at 03:30:15PM +0200, Corinna Vinschen wrote: >>>Urgh. I stumbled over the need_directory flag only two days ago. >>>while debugging the symlink errno problem you reported on the list. >>>CGF is my witness. It's the reason I made the trailing slash change in >>>symlink rather than in path_conv::check. It's quite tricky to keep all >>>possible cases working. Have you tested this change with the entire >>>coreutils testsuite? It seems to be quite thorough. >> >>Yes, I'm a witness. I mentioned that this has to still work: >> >>ls -l foo/ >> >>where foo is a directory. I think that requirement is the cause for > >Actually, where foo is a symlink pointing to an existing directory.
Yes. Imprecise language on my part. >>some complication in the code. >> >>Do we really want to be making many changes to this code at this point >>in the release cycle? Maybe we could branch 1.7.1 and keep making >>riskier changes on the trunk, just like the real projects do it? > >Also less risky would be to make changes locally in mkdir, link, and >rename for now. I'm not clear if this is a regression or not. If it isn't a regression, I'd opt for leaving it until 1.7.2. cgf