hdtodd@eagle.wesleyan.edu (03/01/91)
Is there a sendmail for the Apollo systems (I run 3500/2500, SR10.2) that supports MX records? Has anyone ported newer versions of the code? Is it available at an FTP site? Thanks for any help you can offer. David Todd Wesleyan University
etb@milton.u.washington.edu (Eric Bushnell) (03/01/91)
In article <1991Feb28.173708.39510@eagle.wesleyan.edu> hdtodd@eagle.wesleyan.edu writes: > > Is there a sendmail for the Apollo systems (I run 3500/2500, SR10.2) >that supports MX records? Has anyone ported newer versions of the code? Is it >available at an FTP site? > I compiled sendmail 5.65 obtained from an ftp site at U of Illinois. It's working fine at 10.2 and 10.3. You need to undefine the SETPROCTITLE and compile the getloadavg.c that's supplied as a supplementary file. It's available at many sites. I think I got it from uxc.cso.uiuc.edu, or something like that. Now, has anybody got a version of sendmail that works with BIND on a 10.1 node? Or a version of BIND that works with sendmail? disclaimer: I don't remember the details. Eric Bushnell UW Civil Engr etb@zeus.ce.washington.edu
jf@ap.co.umist.ac.uk (John Forrest) (03/01/91)
In article <17411@milton.u.washington.edu>, etb@milton.u.washington.edu (Eric Bushnell) writes: |> In article <1991Feb28.173708.39510@eagle.wesleyan.edu> hdtodd@eagle.wesleyan.edu writes: |> > |> > Is there a sendmail for the Apollo systems (I run 3500/2500, SR10.2) |> >that supports MX records? Has anyone ported newer versions of the code? Is it |> >available at an FTP site? |> > |> I compiled sendmail 5.65 obtained from an ftp site at U of Illinois. |> It's working fine at 10.2 and 10.3. You need to undefine the SETPROCTITLE |> and compile the getloadavg.c that's supplied as a supplementary file. |> It's available at many sites. I think I got it from uxc.cso.uiuc.edu, |> or something like that. With the latest version from UIUC (5.65a) you don't need to change getloadavg - it uses the xload one. However, there would appear to be a slight bug on 10.2/3 which is currently being worked on. I'd still mount this version though, I expect the fix will be a small patch (they usually are) and generally it seems to work fine. |> |> Now, has anybody got a version of sendmail that works with BIND on a |> 10.1 node? Or a version of BIND that works with sendmail? We run this version (5.65a) on 10.1 - we did the previous IDA version too, although with more problems. We have "upgraded" BIND to get it to work though - we ported BIND 4.6 by turning the resolver into an inlib library, adding it to the system configuration and removing the old resolver entries from the C library. We've still got the code somewhere if someone is interested - it took some rewriting of the source, and we never did get the debug stuff to work properly. However, the resolver works fine, and named did too once you realise that this also suffers from the SETPROCTITLE problem! John Forrest Dept of Computation UMIST
ianh@bhpmrl.oz.au (Ian Hoyle) (03/12/91)
jf@ap.co.umist.ac.uk (John Forrest) writes: >|> I compiled sendmail 5.65 obtained from an ftp site at U of Illinois. >|> It's working fine at 10.2 and 10.3. You need to undefine the SETPROCTITLE >|> and compile the getloadavg.c that's supplied as a supplementary file. >|> It's available at many sites. I think I got it from uxc.cso.uiuc.edu, >|> or something like that. >With the latest version from UIUC (5.65a) you don't need to change getloadavg - >it uses the xload one. However, there would appear to be a slight bug on >10.2/3 which is currently being worked on. I'd still mount this version >though, I expect the fix will be a small patch (they usually are) and >generally it seems to work fine. Paul Pomes from uiuc just wrote to me advising the problem: >The bug was found to be a previous patch that assumed that SIGCLD == SIGCHLD. >When that isn't true (like on Apollos), two different signal handlers are >in a race for the local mail child. > >It now works fine on Apollos. I'm told that /bin/mail does not need to be >replaced provided that the registry daemon is happy and running. The new sendmail version is pub/sendmail-5.65b+IDA-1.4.3.tar.Z available at uxc.cso.uiuc.edu Seems to be rock solid so far on my DN3500 running sr10.3 which acts as our gateway node. ian -- Ian Hoyle /\/\ Image Processing & Data Analysis Group / / /\ BHP Research - Melbourne Laboratories / / / \ 245 Wellington Rd, Mulgrave, 3170 / / / /\ \ AUSTRALIA \ \/ / / / \ / / / Phone : +61-3-560-7066 \/\/\/ FAX : +61-3-561-6709 E-mail : ianh@bhpmrl.oz.au
jf@ap.co.umist.ac.uk (John Forrest) (03/13/91)
In article <ianh.668750023@morgana> ianh@bhpmrl.oz.au (Ian Hoyle) writes: >The new sendmail version is pub/sendmail-5.65b+IDA-1.4.3.tar.Z available at >uxc.cso.uiuc.edu > >Seems to be rock solid so far on my DN3500 running sr10.3 which acts as our >gateway node. > This turned out to be a great problem on our system (10.1 with modified BIND). After some exploration with dde, it turned out the getloadavg.c file had been changed so that the function prototypes were no longer used (for proc1_$get... functions). This was to allow compilation with -N ansi - ie. to allow Apollo compilers to compile under 10.1 - but the resultant binaries fall over (at least on our system), because the function prototypes are essential. The best fix is to return to version 1.4 (use co) and to put the following line in the makefile: CC=cc -U__STDC__ This has worked for us. If anyone else is having trouble, I'd suggest trying this. John Forrest Dept of Computation UMIST PS. I've yet to try the 1.4.3 IDA .cf generation yet, but this is less machine dependent.
jf@ap.co.umist.ac.uk (John Forrest) (03/13/91)
In article <1991Mar12.220948.8204@cns.umist.ac.uk>, jf@ap.co.umist.ac.uk (John Forrest) writes: |> The best fix is to return to version 1.4 (use co) and to put |> the following line in the makefile: |> |> CC=cc -U__STDC__ |> |> This has worked for us. If anyone else is having trouble, I'd |> suggest trying this. I've just realised I forgot to say "v1.4 of getloadavg.c" - the rest is fine. If anyone needs to do this and is still using the version of rcs shipped with 10.1 you will have a problem, because it is not compatible - the simplest way is to edit the RCS/getloadavg.c,v file and remove the line: branch ; To be honest, though, I believe there are still a few 10.1 funnies out there in the binaries. My uk.extras patches contain some. I still believe the conf.h settings: EXPLICIT_CLOSE_ON_EXEC and DOPRNT_FIX are required for 10.1. Unfortunately, the current directory still refers to the previous (5.65a+IDA1.4.2) distribution, and doesn't quite work with this one. I've done the patches so I will soon be putting a new patch directory together. John Forrest Dept of Computation UMIST