warren@ginosko.samsung.com (Warren Lavallee) (07/10/89)
Does anyone have bind 4.8 workin consistently under OSx 4.4c?
I got bind 4.8, then applied a patch from Karl Kleinpaste of Ohio State,
and it still dies with a Bus error every now and then.
I called Pyramid, they said that it is fixed under 5.0. But I
want a working nameserver now...
Any help would be greatly appreciated.
- Warren
Machine Info: OSx32N ginosko 4.4c-890501 0510s Pyramid-9810
--
Your mind understands what you have been taught; your heart, what is true.
NEARnet/Internet: warren@ginosko.samsung.com | SRI-NIC handle: WJL17
US Mail: Samsung Software America. One Corporate Dr. Andover, Ma. 01810
UUCP: {decvax!{gsg,cg-atla},uunet,ulowell}!ginosko!warrencsg@pyramid.pyramid.com (Carl S. Gutekunst) (07/11/89)
>Does anyone have bind 4.8 workin consistently under OSx 4.4c? I called >Pyramid, they said that it is fixed under 5.0. But I want a working >nameserver now... Before I answer this, you need to understand the scope of the problem. When you change versions of BIND, you change getbyhostname(3N) and its friends in libc.a. That means *everything* that references gethostbyname() has to be re- linked and replaced. That works out to a total of 99 binary files in the OSx release, including libc.a itself. You also must relink all your own utilities that call gethostbyname(3N), and only *you* know where they are. You also have to use Sendmail 5.61, of course. That said, there are two versions of BIND 4.8 available from Pyramid for OSx 4.4c. One was done by Field Engineering, and was provided as a special to a customer who was (quite reasonably) threatening to pitch their machine into the Atlantic if we didn't give them a working nameserver. But it was not made available as a general PTF because of the huge number of files affected. The documentation was incomplete at that time, too. The other version was done by me, and is the same code as in OSx 5.0. I have provided this to two sites as private beta tests (both of whom are personal friends, which makes debugging easier), and am also shipping it to a third. Being as I am in R&D, not customer service, I'd really rather not have to support another site. My suggestions are these: - Get yourself high on the priority list for OSx 5.0. This will be one of the best initial releases we have done, if not the best; I would not be afraid to be the first on the block to have it. - Failing that, call up RTOC again, tell them you are perfectly willing to take BIND 4.8 as a "special," and tell your salescritter that if you don't get it you'll pitch the machine into the Pacific. :-) - Failing that, you can get it from me. But if it you have problems with it other than bugs, I won't be able to help you. And if you have some other problem that requires RTOC to service the machine, you'll have to back out all the BIND stuff first. <csg>
csg@pyramid.pyramid.com (Carl S. Gutekunst) (07/11/89)
In article <76769@pyramid.pyramid.com> I wrote: >... and tell your salescritter that if you don't get it you'll pitch the >machine into the Pacific. :-) Oops. Samsung America is in Massachusetts. How far than you throw? :-) <csg>
mb@rex.cs.tulane.edu (Mark Benard) (07/11/89)
In article <76769@pyramid.pyramid.com> csg@pyramid.pyramid.com (Carl S. Gutekunst) writes: >(from warren@ginosko.samsung.com) >>Does anyone have bind 4.8 workin consistently under OSx 4.4c? I called >>Pyramid, they said that it is fixed under 5.0. But I want a working >>nameserver now... I was the one who sent Warren the patch that I got from Karl Kleinpaste. It has worked fine for me. I have not had a named problem in several months. > >Before I answer this, you need to understand the scope of the problem. When >you change versions of BIND, you change getbyhostname(3N) and its friends in >libc.a. That means *everything* that references gethostbyname() has to be re- >linked and replaced. That works out to a total of 99 binary files in the OSx >release, including libc.a itself. You also must relink all your own utilities >that call gethostbyname(3N), and only *you* know where they are. You also have >to use Sendmail 5.61, of course. > 99? I think that is a slight exaggeration. I installed BIND 4.8 and only replaced a handful of files. (However, I admit, it did take me days to get lib.a correct.) And you do not HAVE to run Sendmail 5.61, although that means you cannot fully take advantage of BIND 4.8. I am still running the sendmail distributed with OSx4.4, with a little smail mixed in. But at least I have a working version of named. -- Mark Benard Department of Computer Science INTERNET & BITNET: mb@cs.tulane.edu Tulane University USENET: [{ames,bionet}!]rex!mb New Orleans, LA 70118
csg@pyramid.pyramid.com (Carl S. Gutekunst) (07/11/89)
>>When you change versions of BIND, you change getbyhostname(3N) and its >>friends in libc.a. That means *everything* that references gethostbyname() >>has to be relinked and replaced. That works out to a total of 99 binary >>files in the OSx release, including libc.a itself. > >99? I think that is a slight exaggeration. No sir, that's a 'wc' of my manifest for my BIND 4.8 beta tape. Well, OK, I can root out all the "hangers on," things that come with a package but don't actually call gethostbyname(3N); uux and uucp for example.... 69 files. A slight exaggeration, I guess. :-) >And you do not HAVE to run Sendmail 5.61, although that means you cannot >fully take advantage of BIND 4.8. True; and strictly speaking, you don't *have* to replace any of the routines that call gethostbyname(3N) either. The calls to the nameserver are allegedly upward compatible. But now I'm being pedantic. I guess I think of BIND 4.8 and Sendmail 5.59+ as inseperable; each has features that implies the other. <csg>
warren@ginosko.samsung.com (Warren Lavallee) (07/11/89)
From an article by csg@pyramid.pyramid.com (Carl S. Gutekunst): > strictly speaking, you don't *have* to replace any of the routines > that call gethostbyname(3N) either. The calls to the nameserver are allegedly > upward compatible. But now I'm being pedantic. I guess I think of BIND 4.8 > and Sendmail 5.59+ as inseperable; each has features that implies the other. The support in the library is there and it works. However I did have to link my sendmail v5.61+ (with IDA enchancements) with the libresolv.a library to get it to work. My nameserver hasn't died in 24 hours. This is definately an improvement. I compiled it with -OG... Is that ok? This was a Pyramid C compiler bug that caused this right? Perhaps I won't dump my system into the Atlantic, but I can replace it with a Sequent (All hail Sequent!) Symetry and not upgrade the Pyramid anymore. Who ever heard of booting from Floppy? I do like one thing I saw in the Pyramid. My system disk went off line and the system didn't crash. It was offline for 3 hours and the system wasn't hung. All the access were blocked, but other things worked fine. Like to see a Sequent do that. - Warren -- Your mind understands what you have been taught; your heart, what is true. NEARnet/Internet: warren@ginosko.samsung.com | SRI-NIC handle: WJL17 US Mail: Samsung Software America. One Corporate Dr. Andover, Ma. 01810 UUCP: {decvax!{gsg,cg-atla},uunet,ulowell}!ginosko!warren
csg@pyramid.pyramid.com (Carl S. Gutekunst) (07/12/89)
>I compiled it with -OG... Is that ok? This was a Pyramid C compiler bug >that caused this right? -OG is fine. There was a code generator botch that tripped over the most convoluted ternary expression you've ever seen. This botch occured regardless of the optimizer level. Fixed in OSx 5.0, BTW. >Who ever heard of booting from Floppy? What would you suggest? You can't do the IMPL from hard disk, unless we had a dedicated disk just for the SSP. DEC used an HP catridge, although those proved to be more expensive and less reliable than a simple 5" floppy. The tape is also inappropriate for automatic dumping of hardware diagnostics. In fact, the MIServer uses a RAM card. But those didn't exist a few years ago, at least not ones large enough to be a suitable standin for a disk. The Symmetry doesn't need a similar mechanism because it doesn't have micro- code, and therefore doesn't have the concept of Initial Micro Program Load (IMPL). And it doesn't do hardware error logout to a diagnostic console. <csg>
car@trux.UUCP (Chris Rende) (07/19/89)
In article <76854@pyramid.pyramid.com>, csg@pyramid.pyramid.com (Carl S. Gutekunst) writes: > >Who ever heard of booting from Floppy? > > What would you suggest? You can't do the IMPL from hard disk, unless we had > a dedicated disk just for the SSP. DEC used an HP catridge, although those > proved to be more expensive and less reliable than a simple 5" floppy. The > tape is also inappropriate for automatic dumping of hardware diagnostics. Does the boot process require that floppy to be present? If so it's a scary thought. A chain is only as strong as its weakest link. I'd be very uncomfortable knowing that my mightly mega-$$$$$ Unix box could be rendered useless by a missing or blown 50 cent floppy. car. -- Christopher A. Rende Central Cartage (Nixdorf/Pyramid/SysV/BSD4.3) uunet!edsews!rphroy!trux!car Multics,DTSS,Unix,Shortwave,Scanners,StarTrek ...!sharkey!rphroy!trux!car Minix,PC/XT,Mac+,TRS-80 Model I: Buy Sell Trade "I don't ever remember forgetting anything." - Chris Rende
steve@polyslo.CalPoly.EDU (Steve DeJarnett) (07/20/89)
In article <209@trux.UUCP> car@trux.UUCP (Chris Rende) writes: >In article <76854@pyramid.pyramid.com>, csg@pyramid.pyramid.com (Carl S. Gutekunst) writes: >> >Who ever heard of booting from Floppy? >> What would you suggest? You can't do the IMPL from hard disk, unless we had >> a dedicated disk just for the SSP. > >Does the boot process require that floppy to be present? Yup. >If so it's a scary thought. A chain is only as strong as its weakest link. >I'd be very uncomfortable knowing that my mightly mega-$$$$$ Unix box could >be rendered useless by a missing or blown 50 cent floppy. You should "request" that RTOC send you several COS floppies, then configure them for your system. Keep some in different places so that if you blow one (it happens from time to time), all you have to do is grab a new one and put it in. You're back up in a few minutes. ------------------------------------------------------------------------------- | Steve DeJarnett | Smart Mailers -> steve@polyslo.CalPoly.EDU | | Computer Systems Lab | Dumb Mailers -> ..!ucbvax!voder!polyslo!steve | | Cal Poly State Univ. |------------------------------------------------| | San Luis Obispo, CA 93407 | BITNET = Because Idiots Type NETwork | -------------------------------------------------------------------------------
hoyt@polyslo.CalPoly.EDU (Sir Hoyt) (07/20/89)
In article <209@trux.UUCP> car@trux.UUCP (Chris Rende) writes: >Does the boot process require that floppy to be present? Yup. >If so it's a scary thought. A chain is only as strong as its weakest link. >I'd be very uncomfortable knowing that my mightly mega-$$$$$ Unix box could >be rendered useless by a missing or blown 50 cent floppy. Pyramid is not the first company to use this. There is another very BIG company that use 8" floppies to this day to boot thier machines from. Do you know who?? Can you say 'IBM' ( god that hurts...). The reason I am told that they use them is becuase of reliabilty! -- John H. Pochmara A career is great, UUCP: {csun,voder,trwind}!polyslo!hoyt But you can't run your Internet: hoyt@polyslo.CalPoly.EDU fingers through its hair -Graffiti 4/13/83
csg@pyramid.pyramid.com (Carl S. Gutekunst) (07/21/89)
>Does the boot process require that floppy to be present? >If so it's a scary thought. A chain is only as strong as its weakest link. Like I said, feel free to suggest something better! No other media that could store 300K of data and that was available in 1985 was as reliable as a plain ol' 5" floppy. The floppies do die from time to time (of the five machines I use, we've lost one in the past two years); so you keep extras. No problem. The MIServer uses a RAM catridge -- a technology unavailable in 1985 -- mostly because some customers complained about the floppies. Not because the floppy didn't work, or wasn't reliable, but because they were squeemish about doing IMPL from a floppy. Always happy to please, Pyramid got rid of the floppies. Of course, the RAM catridges cost $80 each, while a floppy costs 50 cents. <csg>