jsol@bu-cs.BU.EDU (Jon Solomon) (03/09/89)
Does anyone have a copy of named 4.8 (it must be 4.8) for the encore? --jsol
clyde@ut-emx.UUCP (Clyde W. Hoover) (03/09/89)
We run BIND 4.8 (and resolver) on under UMAX 4.2 (rev 3.2) send mail to address below for more information (its not really a port, but there were some changes made). -Clyde Hoover Shouter-To-Dead-Parrots @ Univ. of Texas Computation Center; Austin, Texas clyde@emx.utexas.edu; ...!cs.utexas.edu!ut-emx!clyde "You really have to take a broad perspective when giving pat answers to other people's problems." - Eyebeam
loverso@Xylogics.COM (John Robert LoVerso) (03/10/89)
In article <28537@bu-cs.BU.EDU> jsol@bu-cs.BU.EDU (Jon Solomon) writes: > Does anyone have a copy of named 4.8 (it must be 4.8) for the encore? Bind4.8 ports readily to UMAX4.2. The version we're using here (Xylogics) on our MM310 is actually the version from the "redistributable" BSD sources available on uunet.uu.net. That version, among other things, has had some bugfixes applied to it over the original Bind4.8 distribution (Mike Karels posted at least 3 bugfixes). For the most part, I'm not running with any other changes other than these two: In named/ns_main.c, "#ifdef notdef" the SIOCGIFADDR ioctl. It doesn't work in UMAX4.2 R3.2 [I'm not sure why and I no longer have access to kernel source], but thats ok because the SIOCGIFFLAGS directly above it returns the interface address, too. In tools/nslookup/res.h, add the lines #ifdef NAME_LEN #undef NAME_LEN #endif directly before the #define of NAME_LEN. This is because some include file in /usr/include (I forget which one) defines NAME_LEN, and so they conflict. In addition to that, I am also using the cleaver trick Stuart Levy mailed to the bind list that allows the 4.3 struct hostent to be binary compatible with the 4.2 one; i.e., my netdb.h has: struct hostent { char *h_name; /* official name of host */ char **h_aliases; /* alias list */ int h_addrtype; /* host address type */ int h_length; /* length of address */ #define h_addr h_addr_0 /* Dummy to make resolver think we're 4.3 */ char *h_addr_0; /* This field for 4.2 binary compatibility. */ char **h_addr_list; /* list of addresses from name server */ }; and change res/named/gethostnamadr.c, around line 120, to: #if BSD >= 43 || defined(h_addr) /* new-style hostent structure */ *hap = NULL; #endif host.h_addr = h_addr_ptrs[0]; return (&host); This change means that you can still put the bind netdb.h in /usr/include, and leave the original Encore 4.2 code libc.a. Someone linking against the old gethostbyname & friends with the new netdb.h will still work. But, I've just got a libresolv.a contains the new gethostbyname & friends (actually, it has all the 4.3tahoe src/lib/libc/net code). While I was at it, I also added in some additions from 4.3tahoe libc, such as strcasecmp and setenv. Good news: Encore's "4.3" os (UMAX R4.0) will have Bind. Bad news: Last I heard, they were going with Bind 4.3 (off the 4.3 tapes). (sigh) John -- John Robert LoVerso, Annex Software Engineer Xylogics, Inc., 617/272-8140 [formerly with Encore Computer Corp] encore!xylogics!loverso, loverso%Xylogics.COM@Encore.COM
rww@ibuki.UUCP (Richard Weyhrauch) (03/11/89)
d 102