craig@scorplx.UUCP (Craig Miller) (09/21/84)
#
I have seen these problems in both the 2.10 beta and the 2.10.2
distributions. If you are running into one or both of these problems,
please send me mail indicating the problem(s) and what kind of system
you are running on. If I get enough responses, maybe I can come up with
some kind of pattern (i.e. 68000 machines...) and see about asking the
people in charge to add some #ifdef's in...
1) assuming that a setuid program on a USG system can setuid()
back after doing a setuid(getuid())... (is this just in SysV,
or am I the only SysIII site that can't do this?) The function
involved is mknewsg() in "inews.c"...
2) the name 'sigtrap' (an integer used globally in a few files)
is also the name of a function in /lib/libc.a (if I remember
right, I kept getting bus errors until I found the problem
and fixed it)
Send me mail if you come across either of these things (or if you just
have information on them) ... Thanks.
--
Craig Miller
{decvax,harvard}!scorplx!craig
Honeywell Info. Systems, Billerica, Mass.craig@scorplx.UUCP (Craig Miller) (10/10/84)
/* ... */ (If you were waiting for a summary on my questions, sorry for taking so long... I tried to wait till I was sure all of the responses were in... I guess maybe I waited a bit too long...) Anyway, the problems were: > 1) assuming that a setuid program on a USG system can setuid() > back after doing a setuid(getuid())... (is this just in SysV, > or am I the only SysIII site that can't do this?) The function > involved is mknewsg() in "inews.c"... There were only a few people who replied to this one... According to them, only SysV can setuid back. One person said they thought only SysV release 2 could do this. Apparently we SysIII sites are out of luck... Any suggestions on what to do about this? Maybe instead of #ifdef USG, someone should consider #ifdef SYS3, #ifdef SYSV (or something similiar). Comments? > 2) the name 'sigtrap' (an integer used globally in a few files) > is also the name of a function in /lib/libc.a (if I remember > right, I kept getting bus errors until I found the problem > and fixed it) I received numerous answers to this one... (mostly of the format: "I have machine X running version Y...") The only thing common (that I could tell) was that all machines with this problem are 68000 machines... The other thing mentioned by more than one person was Unisoft. (hey Unisoft, did you guys add 'sigtrap' to the C library - in particular, something to do with signal(2)?) If Unisoft is the source of this, maybe #ifdef UNIPLUS, or something similiar should be added... I am still not sure... Any more help would be greatly appreciated... Got any suggestions, Rick (Adams)? Anyway, thanks to all who responded... -- Craig Miller Honeywell Info. Systems, Billerica, Mass. {decvax,harvard}!scorplx!craig Any spelling or grammatical errors were (mis)produced by the mispelling daemon...
karsh@geowhiz.UUCP (Bruce Karsh) (10/13/84)
> > Anyway, the problems were: > > > 1) assuming that a setuid program on a USG system can setuid() > > back after doing a setuid(getuid())... (is this just in SysV, > > or am I the only SysIII site that can't do this?) The function > > involved is mknewsg() in "inews.c"... > > There were only a few people who replied to this one... According to > them, only SysV can setuid back. One person said they thought only SysV > release 2 could do this. Apparently we SysIII sites are out of luck... > Any suggestions on what to do about this? Maybe instead of #ifdef USG, > someone should consider #ifdef SYS3, #ifdef SYSV (or something similiar). > Comments? > I didn't really want to respond to this one because I am pretty new at UUCP and news, but since nobody else did, here goes... We had some problems in inews.c on our system. They had to do with permissions on files. I believe that we had the problems that you are referering to here. What i did is to take the section in vnews where the setuid(getuid()) occurs and put it into a fork. In addition, I experienced sporadic file linking failures. So I added some code to log them more carefully and to retry once. So far, I've had no cases where more than one retry is required. If anyone out there in netland can explain why that is necessary, I'd like to hear from you. We are running on a Masscomp system. I think it was based on system III, but I think it has a lot of BSD4.2 stuff in it. Here are the diffs: 470,482d469 < /* Debugging code added by BK at UW geowhiz */ < else < { < logerr("Failed once %s\n",ARTICLE); < e = errno; /* keep log from clobbering it */ < logerr("Error code is %d",e); < if(link(ARTICLE,bfr) == 0) < { < logerr("Succeeded on second try\n",ARTICLE); < break; < } < logerr("Failed twice %s\n",ARTICLE); < } 486,489d472 < /* Debugging code added by BK at UW geowhiz */ < logerr("Error code is %d",e); < logerr("ARTICLE is %s",ARTICLE); < logerr("bfr is %s",bfr); 794,796d776 < /* Next two lines added by BK at UW geophysics */ < if((pid-fork())<=0) < { 808,812d787 < /* Next four lines added by BK at UW geophysics */ < exit(0); < } < else < fwait(pid);
lwall@sdcrdcf.UUCP (Larry Wall) (10/23/84)
In article <1469@pur-phy.UUCP> crl@pur-phy.UUCP (Charles LaBrec) writes: >Now users' "options" lines are disappearing with 2.10.2. We are >on a PDP-11/44 with 2.8 BSD. Was this version really tested? There >seem to be a number of serious problems. I've already posted others, >like non-control messages in control, lots of articles with no titles, >parts of articles being pre-pended to .newsrc's, and empty articles >(no header even) in junk! > Mea culpa, but only on the empty junk articles. See rn bug #17 (actually a problem with the inews patch). Larry Wall {allegra,burdvax,cbosgd,hplabs,ihnp4,sdcsvax}!sdcrdcf!lwall
steiny@scc.UUCP (Don Steiny) (10/28/84)
> Now users' "options" lines are disappearing with 2.10.2.
The code for SORTACTIVE was eating .newsrc files here
(on an Interdata 8/32 with Wollongong V7). I don't have
time to figure out why and it is not obvious, so I undefined
the SORTACTIVE code and have had no problems since.
--
scc!steiny
Don Steiny - Personetics @ (408) 425-0382
109 Torrey Pine Terr.
Santa Cruz, Calif. 95060
ihnp4!pesnta -\
fortune!idsvax -> scc!steiny
ucbvax!twg -/crl@pur-phy.UUCP (Charles LaBrec) (11/01/84)
Now users' "options" lines are disappearing with 2.10.2. We are on a PDP-11/44 with 2.8 BSD. Was this version really tested? There seem to be a number of serious problems. I've already posted others, like non-control messages in control, lots of articles with no titles, parts of articles being pre-pended to .newsrc's, and empty articles (no header even) in junk! Charles LaBrec UUCP: pur-ee!Physics:crl, purdue!Physics:crl INTERNET: crl @ pur-phy.UUCP