On Tue, 13 Sep 2011, Guillem Jover wrote: > On Mon, 2011-09-12 at 08:25:13 +0000, Raphaël Hertzog wrote: > > The following commit has been merged in the master branch: > > commit 9dcb50a59414e1fe0e6a8a4c8ee9f59d10706396 > > Author: Raphaël Hertzog <[email protected]> > > Date: Mon Sep 12 09:42:11 2011 +0200 > > > > update-alternatives: new non-regression test > > > > This one catches the regression introduced in > > 6c481e19a532080643e3cf5c76760df8009ea2c7 and fixed in > > 9dceac005a0331ceb840f9c8273c7fd195413516. > > > > It should also catch the former problem of passing NULL to stat() (via > > new_file). But this did not trigger any segfault, merely a failure > > with EFAULT. > > Hrmm, should not have trusted the existing code when changing to > xreadlink. Because even if on Linux (and other kernels) it returns > EFAULT, due to the argument being handled directly by the kernel, > there's no guarantee it cannot segfault, and the system has all the > right to do so, as such that's not a portable assumption to make. > For example it *does* segfault on the Hurd.
Yeah, there definitely was a theoretical bug there. But in practice it's almost never triggered as the alternative link usually only exists if the corresponding /etc/alternatives link is there too. Cheers, -- Raphaël Hertzog ◈ Debian Developer Follow my Debian News ▶ http://RaphaelHertzog.com (English) ▶ http://RaphaelHertzog.fr (Français) -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/[email protected]
