tobeye@northstar.dartmouth.edu (Anthony Edwards) (03/01/90)
Perhaps you can help me. I installed ATK4 from the X11 release tape and everything worked fine. I applied patches 1 and 2 and things were still honky-dorry. Then, I applied the latest patch (we're at 3 now, right?) and now Messages doesn't recognize that our system uses the Name Resolver for mail address. I have verified that NAME_RESOLVER in the ./config directory IS defined and never undefined. Perhaps a clue: before this problem arose, I discovered an error in the compile whereby the symbol dn_skip() came up undefined. I back-tracked this problem to a file in the ./overhead/mail/lib directory whereby I needed to force DN_SKIP to be dn_skipname() rather than dn_skip(). Why I think the NAME_RESOLVER is the problem: When I send a message to a foreign host, even one that ends in .edu, I must verify that that hosts exists. I definitely did not have to do this before. Local hosts that I know are in our /etc/hosts file never cause problems. Any ideas? Can I give you more info? - Anthony
tobeye@NORTHSTAR.DARTMOUTH.EDU (Anthony Edwards) (03/01/90)
It turns out that I do have the resolver turned ON. Somehow, I seemed to have removed a line from the AndrewSetup file: AMS_ValidDomainSuffixes: .edu, .com, .bitnet which supresses the warning message about hosts not existing if the host name had one of these suffixes. I played with the test program that Craig suggested I compile and run. It mostly gave me the results that a host was "indeterminate" which on the program's yes-maybe-no answer scale is a maybe. I thought name resolution was more accurate than that, but I guess not. For those interested in seeing the ouput from the valhost.test program, here it is: Test your favorite destination mail domain names. One at a time, please. domain: transarc.com Input: ``transarc.com''; output: ``'' (indeterminate) domain: hinman Input: ``hinman''; output: ``'' (indeterminate) domain: ecs.umass.edu Input: ``ecs.umass.edu''; output: ``umass.edu'' (indeterminate) domain: dartmouth.edu Input: ``dartmouth.edu''; output: ``'' (indeterminate) domain: andrew.cmu.edu Input: ``andrew.cmu.edu''; output: ``andrew.cmu.edu'' (good) domain: northstar Input: ``northstar''; output: ``northstar.dartmouth.edu'' (good)
tobeye@NORTHSTAR.DARTMOUTH.EDU (Anthony Edwards) (03/02/90)
> Now, we did work around a bug in the RT's hc compiler (hc 1.4r) that > was causing bogus names (that it should have rejected) to come out as > ``indeterminate''. Might that be a possible reason why you get some of > your ``indeterminate'' answers? It very well could be. I used hc1.4 to compile the whole ATK. > And do you have the HardHostValidation flag off, so that all hard failures > (``bad'') map to soft ones (``indeterminate'')? (This is a fix not in any > official patch, yet.) I put AMS_HardHostValidationErrors: no into our AndrewSetup file which explains why my errors were softened. I'm really confused why Messages can't pickup on the use of the Name Resolver. I know we have it installed and I thought I've been using it in other programs for some time now (telnet, finger, etc). Just to verify that ATK is square on the Name Resolver, I checked: % cd ${ANDREWDIR}/config % grep RESOLVER * allsys.h:#define RESOLVER_ENV 1 allsys.mcr:#ifdef RESOLVER_ENV allsys.mcr:#else RESOLVER_ENV allsys.mcr:#endif RESOLVER_ENV allsys.mcr.orig:#ifdef RESOLVER_ENV allsys.mcr.orig:#else RESOLVER_ENV allsys.mcr.orig:#endif RESOLVER_ENV To make absolutely sure that this was really affecting compiles such as valhost.test, I issued a cc -E and then searched for text wich valhost.c will only include if RESOLVER_ENV is turned on. I found the text. The only thing I can conclude is: a) our system administrator mucked with my machine and accidentally turned off the name resolver. (Do you know of a quick, non-ATK, way of checking for the Name Resolver?) b) ATK is somehow, someway, subtly not connecting to the Name Resolver library routines. - Anthony