On Sat, Jun 03, 2006 at 03:49:42PM -0500, PauAmma wrote: > On Sat, 3 Jun 2006, Kris Kennaway wrote: > > >On Fri, Jun 02, 2006 at 10:45:23PM +0200, Ulrich Spoerlein wrote: > >>PauAmma wrote: > >>> > >>>(http://www.freebsd.org/cgi/query-pr.cgi?pr=76915 for details) > >> > >>I didn't read the patch (yet) and I'm not familiar with the code, but if > >>we would enforce all pkg-install/deinstall scripts to be sh(1) scripts, > >>then it's simply a matter of replacing ./foo with 'sh foo', right? > > That would require changes to pkg_delete and related utilities, which > may actually be a good idea (see below). > > >And potentially many other similar changes. The full scope of the > >changes required to fully support a noexec /var is clearly enormous, > >which is one reason why I don't want to add partial support for this > >nonstandard and rarely-used configuration. > > IMO this wouldn't be supporting it strictly speaking, more along the lines > of not trying to and failing messily. That said, I'm beginning to think > that if this specific problem should be fixed at all, the fix needs to go > in pkg_delete and friends, not the ports system itself: > > 1- This would address the performance concerns Sergey Matveychuk raised. > > 2- The same problem happens when using pkg_delete -f directly. > > Considering the above, and so I don't embark onto something else that only > I care about, where would you suggest I ask before I start? > > (Oh, and since this is no longer ports-related, you can probably close the > PR unless you think someone may want to revisit it later.) Personally I don't think it is worth attempting to handle, as I mentioned previously. Kris
Attachment:
pgpkdmvoTSt9l.pgp
Description: PGP signature