On Wed, 28.01.15 14:25, Martin Pitt (martin.p...@ubuntu.com) wrote: > Hello Lennart, > > Lennart Poettering [2015-01-28 14:19 +0100]: > > I really don't follow. Why do we need the add_alias() equality check > > now? I mean, if sysv_translate_facility() now detects if we translate > > facilities listed in Provides: as pointing to the script itself, and > > handles .sh properly, why do we need to check in add_alias() a second > > time if we try to create a symlink to oursevlves? > > > > The latter can go no, can't it? > > This should indeed not happen now with the recent fixes. This was > merely a defensive check if this situation happens with some other > corner case after committing 29e0e6d8c. But the later commit 77354c is > a stronger hammer for this, so this check is now indeed not that > important any more. > > Chances are that the TOCTOU unlink will be rewritten at some point to > what you suggested (collecting the aliases in memory and writing them > after synthesizing all the units), at which point the defensiveness > might be a tad more important again, but at least now we have a rather > comprehensive test suite which ought to catch all this (like the > regression in 4e48855534 ☺ ). > > So in short: if you don't like it, I'm happy to remove this.
OK, removed it now. Tried to be more careful this time, and "make check" still works. Lennart -- Lennart Poettering, Red Hat _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel