Path: utzoo!attcan!uunet!tut.cis.ohio-state.edu!ucbvax!CAEN.ENGIN.UMICH.EDU!markg From: markg@CAEN.ENGIN.UMICH.EDU (Mark Giuffrida) Newsgroups: comp.sys.apollo Subject: Re: Upgrading to SR10.2 Message-ID: <490382469.0017b5e@caen.engin.umich.edu> Date: 5 Mar 90 15:22:17 GMT Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 50 /usr/new/lib/mh/spost -library //foo/users/awhitton/Mail -verbose \ -watch //foo/users/awhitton/Mail/drafts/8 Segmentation fault >tb Process 1278 (parent 1276, group 1278) Time 90/03/02.07:47(EST) Program /bsd4.3/usr/lib/sendmail Status 00040004: reference to illegal address (OS/MST manager) In routine "bcopy" line 167 Called from "rca_$use_known_rgy" line 1056 Called from "rca_$find_a_candidate_registry" line 1229 Called from "rca_$check_binding" line 1312 Called from "getpwent" line 129 Called from "rgy_unix_$getpwnam" line 214 Called from "rgyc_unix_$getpwnam" line 4409 Called from "getpwnam" line 213 Called from "username" line 393 Called from "setsender" line 443 Called from "main" line 707 Called from "unix_$main" line 114 Called from "" line 31999 Called from "PM_$CALL" line 176 Called from "pgm_$load_run" line 891 Called from "pgm_$invoke_uid_pn" line 1112 This is great. I thought that we were the only ones getting this because of our large registry size. Let me explain what I have reported to apollo so far. I open a call about a month ago on this one (#A2004386). I gave them exactly the same scenario and the exact same traceback. The problem was that I could not get them to reproduce it. I might add at first they said there wasn't much they could do since it happened with "unsupported software" (i.e., the MH). I did manage to get them to file an apr on the problem (apr DDE11). This is a strange bug, because getpwnam() functions correctly 99.99% of the time (except when the registry is loaded and it falsely returns "unknown user", but that is another problem - and another filed apr). When getpwnam() fails under these conditions, it consistantly fails. I tried unsuccessfully to reproduce the problem. I have always felt that this was a combination registry and fork() problem. I hope Apollos boosts the priority of this problem as other sites are now experiencing it. This problem was not there in sr10.1. It also cripples any site dependent on using MH like we are. BTW, I was able to get a workaround by compiling a shared library version of MH. For some reason getpwnam() is stable in that scenario. Mark Giuffrida University of Michigan markg@caen.engin.umich.edu