On Fri, Dec 15, 2006 at 09:05:35PM +0100, Tim Dijkstra wrote: > On Fri, 15 Dec 2006 18:11:21 +0100 > Bill Allombert <[EMAIL PROTECTED]> wrote: > > > So far none of the people affected has been positive that it actually > > fixed the bug for them. This is a problem. > > Sure it does, it fixes it for me. And I think I convincingly argued why > it does. Mario and Yan could you please verufy that the attached patch > works for you to?
Then that's fine with me. You did not say so before and Mario stated he could not test whether it worked. > > What worry me about the patch is the fact that create_lock() and > > check_dpkglock() are not performed in the same order. In particular, if > > create_lock() fail we exit with error 0 instead of 1 thus maybe > > concealing a real error. > > You used to exit with 1 only if both creat_lock and check_dpkglock were > false at the same time. If that is what you want, that is easy to fix. > New patch attached. The locking logic is now the same as in the original. If I am not mistaken, your new patch adds a race condition between create_lock() and check_dpkglock(). If dpkg release the lock after create_lock() and before check_dpkglock(), then we do exit(1) while we should not. In the original version, check_dpkglock was checked first so this race condition was not possible. Sorry to be so slow dealing with this issue... Cheers, -- Bill. <[EMAIL PROTECTED]> Imagine a large blue swirl here. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
