On Friday, February 22, 2019 at 11:23:10 AM UTC-5, Jon Zeppieri wrote: > > > On Fri, Feb 22, 2019 at 10:44 AM Brian Adkins <lojic...@gmail.com > <javascript:>> wrote: > >> >> Yes, I think we found the problem: >> >> $ ls -l /etc/localtime >> lrwxrwxrwx 1 root root 36 Feb 21 21:45 /etc/localtime -> >> /usr/share/zoneinfo/America/New_York >> deploy@ip-172-31-10-34:~$ ls -l /usr/share/zoneinfo/America/New_York >> lrwxrwxrwx 1 root root 13 Jan 1 00:05 >> /usr/share/zoneinfo/America/New_York -> ../posixrules >> >> >> $ racket >> Welcome to Racket v7.1. >> > (require racket/enter) >> > (require tzinfo/private/zoneinfo) >> > (enter! tzinfo/private/zoneinfo) >> <pkgs>/tzinfo/tzinfo/private/zoneinfo> (define zi (make-zoneinfo-source)) >> <pkgs>/tzinfo/tzinfo/private/zoneinfo> (tzinfo-has-tzid? zi >> "America/New_York") >> #t >> <pkgs>/tzinfo/tzinfo/private/zoneinfo> (find-zoneinfo-directory >> default-zoneinfo-search-path) >> "/usr/share/zoneinfo" >> <pkgs>/tzinfo/tzinfo/private/zoneinfo> (detect-tzid/unix (zoneinfo-dir zi) >> >> (find-zoneinfo-directory default-zoneinfo-search-path) >> >> (tzinfo->all-tzids zi)) >> "posixrules" >> >> > Well, that's one I haven't seen before -- a file that actually names a > time zone is a symlink to another file. (I see we're not the only people > who have run afoul of this [ > https://github.com/HowardHinnant/date/issues/252] > <https://www.google.com/url?q=https%3A%2F%2Fgithub.com%2FHowardHinnant%2Fdate%2Fissues%2F252%255D&sa=D&sntz=1&usg=AFQjCNEAZDH1-NNFyKjuxBMiG3wRHM8IDw> > ). > > We have a few options here. > - Simply putting the /etc/timezone check before the /etc/localtime check > would solve the problem for you, though not for systems that use > /etc/localtime as a symlink but do not use /etc/timezone. > - Similarly, not short-circuiting the various checks and only looking for > valid values after we've accumulated all the results of the tests would > work in your case, assuming that (tzid-from-/etc/timezone) returns the > right answer for you (which I assume it does). > - Changing the implementation of the /etc/localtime check to use > readlink(2) to follow the symlink a single step is an option, though it > wouldn't help if someone linked directly from /etc/localtime to posixrules. > That sounds like a crazy scenario, but I would have said that linking > America/New_York to posixrules is unlikely. (I'd expect it to be the other > way around.) I guess the more general approach would be to use readlink(2) > until we find a path that names a time zone. > - I could also abandon the symlink check altogether and always use the > slow path, which checks for file _content_ identity between /etc/timezone > and any file that names an IANA time zone in the zoneinfo tree. > > (Of course, all of this would be unnecessary if the tzid were actually > somewhere _in_ the tzfile(5) format somewhere.) > > If you need an immediate workaround, you can launch racket with the TZ > environment variable set to the contents of /etc/timezone. But I'll fix > this just as soon as I decide what the best approach is. > > > By the way (and this question is for everyone), am I right that > readlink(2) functionality isn't already in the Racket standard library > anywhere? (I don't know if it has an analogue on Windows.) > > - Jon > > Yes, (tzid-from-/etc/timezone) returns the correct value:
<pkgs>/tzinfo/tzinfo/private/os/unix> (tzid-from-/etc/timezone) "America/New_York" It seems that not short circuiting would be a good idea regardless of other changes. It's not urgent for me, because the code in question won't run late in the evening where the problem occurs. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
