Hi Mark, Now I see your points. So IPv4 addresses should have a higher priority over IPv6 when returned by USOCKET host resolving functions. I’m a little busy in this week, *next* week I’ll revise all related USOCKET functions and make the changes (also need a better way to detect IPv6 addresses in various forms). Hopefully at the end of July I can make a new release with cumulative updates in the past... more than one year.
—Chun > Il giorno 04 lug 2018, alle ore 23:44, Mark H. David <m...@yv.org> ha scritto: > > Hi, > Kind of asking for a friend. I thought the thing to do would be to loop > through the list and choose the first non-ipv6 result whereas now it simply > chooses the first result. For the random host case, it should similarly > filter out ipv6 entries in the result. > Thank you very much! > -Mark > > ----- Original message ----- > From: "Chun Tian (binghe)" <binghe.l...@gmail.com> > To: "Mark H. David" <m...@yv.org> > Cc: "usocket-devel" <usocket-devel@common-lisp.net> > Subject: Re: bug in host-to-vector-quad on lispworks (also > host-to-vector-quad) > Date: Wed, 4 Jul 2018 22:47:04 +0200 > > 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 >> > > Email had 1 attachment: > + signature.asc > 1k (application/pgp-signature)
signature.asc
Description: Message signed with OpenPGP using GPGMail
