Vincent Lefevre wrote in <20180821230229.ga16...@zira.vinc17.org>: |On 2018-08-21 07:07:15 -0700, Kevin J. McCarthy wrote: |> I don't like removing 20-year old safety checks, but I think it's okay |> to do so for the case where link() returns 0. | |I was also hesitant, but after thinking more about it: | |* That's an unusual "safety check" (I doubt other software does | the same kind of check after link() with return value 0).
Quite the opposite is true. The original BSD code for creating dotlock files does this, always. |* It is not clear whether this has ever been usuful (nothing in | comment or in the commit log). According to some statement on a list you are also subscribed to this NFS problem has been fixed "at the early 90s". The BSD code i refer to came up later as far as i now (without having digged to thoroughly). |* It might even have been a mistake. I suspect that the goal was | for NFS, but then the test should have been for non-zero return | values instead of zero. The code i refer to was not about renaming some file, but about savely creating a temporary dotlock file, of course, where several processes may create the very same file, and it must be ensured that it is you who won the race. Since i have no overview over Unix systems but *BSD and Linux (2.0.x, 2.2.x, 2.4.x, then a decade not, then 4.x), and Solaris via OpenCSW.org, and an old UnixWare VM but which i have not used since my old machine died, i would not dare to change this. Maybe Christos Zoulas of NetBSD would know more, but then i think i am not the right one to ask. I for one will not change the code for the MUA i maintain, because one of the (very far) future goals is (at least partial) support of elder Unix from before Y2K (from before 1999-01-11, when i first used Linux, to be exact). --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt)
