Hi Mark, thanks for reporting this issue.
It’s obvious that #(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1) means IPv6 address “::1”, which is indeed a valid address mapping to “localhost”. The function HOST-TO-HBO is not an exported function, it’s used internally by some backends and deals only with integers representing IPv4 addresses. But on LispWorks (>=6.1) the function lw-hbo-to-vector-quad may pass longer vectors back to HOST-TO-HBO and then caused this issue. On my Mac OS X laptop, I also got the same: (the first returned address is “::1”) CL-USER 5 > (comm:get-host-entry "localhost" :fields '(:addresses)) (#<COMM:IPV6-ADDRESS ::1 4020187E73> 2130706433) So what’s your desired behavior here? In the quick fixes, I could do something to make get-random-host-by-name return a random IPv4 address in presence of mixed IPv4/v6 address lists. But generally speaking get-hosts-by-name should support IPv6 addresses, especially for those IPv6-only hostnames. On the other side, HOST-TO-HBO is not an exported function, and it’s also not called by LispWorks backends, I don’t understand how your call chain finally touched it and let LispWorks throw an exception. It’s hard to support IPv6 in HOST-TO-HBO. Let me know what you think here, and with your help/inputs we can make usocket better. Regards, Chun > Il giorno 04 lug 2018, alle ore 21:04, Mark H. David <m...@yv.org> ha scritto: > > Running usocket-0.7.0.1 in LispWorks 7.1.0. In my system it happens currently > that > > (get-hosts-by-name "localhost") > > returns the following list: > > (#(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1) #(127 0 0 1)) > > Then get-host-by-name simply takes the first of that list and returns it. > > Host-to-hbo expects the result of get-host-by-name to always be a vector > quad, so the result of calling it with "localhost" is an error on the > recursive call with the result of (get-host-by-name "localhost") => #(0 0 0 0 > 0 0 0 0 0 0 0 0 0 0 0 1) > > The error message is due to falling out of the ecase: > > #(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1) fell through ETYPECASE expression inside > USOCKET::HOST-TO-HBO. > > I observed that host-to-vector-quad gives the wrong result, a 16-byte instead > of a 4-byte vector, due to the same root cause, but it will happen > "randomly". The reason is it uses get-random-host-by-name, which chooses a > random element of the same result list returned by (get-hosts-by-name > "localhost"). For example, I just ran this in the REPL -- log: > > USOCKET> (loop repeat 10 for r = (host-to-vector-quad "localhost") when (not > (= (length r) 4)) collect r) > (#(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1) #(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1) #(0 0 > 0 0 0 0 0 0 0 0 0 0 0 0 0 1)) > > This seems to be LispWorks only. I tried this on SBCL and CCL, but I could > not recreate similar problems. Also, not all MacOS installs seem to get this > weird ipv6 localhost the way I do. I'm not sure why I'm getting this. I do > not remember doing anything special or any kind of stuff with ipv6, but > apparently something I've done has "infected" me with it. However, I can say > I do not notice other networking issues on my system. > > Thanks, > Mark >
signature.asc
Description: Message signed with OpenPGP using GPGMail
