To: siemsen@ucar.edu (Pete Siemsen) cc: Gated-alpha@gated.merit.edu Subject: Re: gated for IRIX 6.2? Date: Tue, 16 Jul 1996 15:12:10 -0700 From: "Steve M. Acheson" I've cc'd this to the list in case anyone else is interested. Having poked through the source, grubbed though the /usr/include files, and tried many simple attempts to trick gated and irix6.2 to like each other, the conclusion is that a port is needed. The methods of accessing the kernel routing table under Irix6.2 is slightly different than anything else. I will be undertaking this port as I can, and would be happy to share with anyone interested, but can't promise any timeframe, etc. For those of technical interest: The problem seems to be that sgi has migrated to the routing socket for updates, etc, but has changed the sockaddr and rtentry significantly. There is no sa_len, or rt_dst, etc. The 4.4 route.h seems to be mostly there, but some sgi specific #defs exist. The socket.h isn't quite what I expected either. While not insurmountable, it's not trivial for me either. And that's just the changes that I've discovered so far. Steve -- ====================================================== Steve Acheson 415-604-4495 Sterling Software sma@nas.nasa.gov NASA Ames Moffett Field, Ca 94035-1000 To: siemsen@ucar.edu (Pete Siemsen) Subject: Re: gated for IRIX 6.2? Date: Tue, 16 Jul 1996 15:56:14 -0700 From: "Steve M. Acheson" Sigh. It seems again that vern is ahead of us or know's more than I do (which is of course the truth) but I've not been able to compile using the patches. The Config file wants sysctl and 6.2 (at least my version) doesn't have it. This makes for difficulty getting the routes, etc. I'm going to play with the options, then go back to porting using vern's patches as pointers... Satch To: siemsen@ucar.edu cc: bjuliano@nas.nasa.gov, kml@nas.nasa.gov, templin@nas.nasa.gov Subject: Gated on 6.2 Date: Tue, 16 Jul 1996 19:24:38 -0700 From: "Steve M. Acheson" Ok, I've figured out what the reason is behind the diffs that Vern set out for the 6.2 gated. There are required patches available from SGI for 6.2 that do things like implement new library calls (sysctl) etc, that make gated work. I'll figure out which ones are required to get gated up and send you an email if you want. That'll be tomorrow though. Satch Date: Wed, 14 Aug 1996 17:05:09 -0600 From: vjs@mica.denver.sgi.com (Vernon Schryver) To: siemsen@ucar.edu (Pete Siemsen) Subject: Re: gated 3.6Alpha2 and IRIX 6.2 > When I upgraded my SGI workstation from 5.3 to 6.2, I lost gated. I > posted to the gated newsgroup asking about gated under IRIX 6.2. > Steve Acheson (sma@nas.nasa.gov) responded with a pointer to your > gated hacks, and the IRIX patches that were needed. I installed the > patches and the hacks, and now have gated running. Thanks!!! > > The question is, your hacks were for gated 3.6Alpha1. I installed > gated 3.6Alpha2. I know it's risky replacing 3.6Alpha2 source files > with "fixed" 3.6Alpha1 files, but I did, and it worked. I'm about to > go back and do it right, diff'ing your files and the 3.6Alpha1 > versions, and editing the differences into the 3.6Alpha2 versions. > > Anyway, the obvious question is, have you informed the Gated > Consortium about your mods? Should I? Perhaps gated 3.6Alpha3 can > include your work, and I won't have to edit future versions... > > Thanks again! You're welcome. I'm discouraged about the Gated Consortium. I had hoped the switch to Merit would fix at least some of the Cornell problems. But no such luck. - when I looked at the new gated's, the configuration stuff in the then current beta of 3.5 was more messed up than in 3.6Alpha1, so I ported 3.6Alpha1. I sent the changes necessary for IRIX 6.2, including a simple bug affecting all systems, to the Consortium months before they announced 3.6Alpha2, and got an acknowledgement. But none of the changes I suggested are in 3.6Alpha2. - one of the reasons SGI has not moved to a newer gated for half a dozen years is that every release of the code is more ballooned than the previous, and at least 30% of the lines have changed. Compare a few of the 9 MByte of 3.6Alpha2 with 3.6Alpha1, and notice they've changed much of it, mostly gratuitously, and often incorrectly. At least I assume they would not intend comments of the form: /* * a * word * per * line */ - the description of 3.6Alpha2's bugs and changes is not encouraging. - the current 3.5 beta has some of the same bugs. When I get back to it, I'm going to replace that file you found with a similar file, but for the 3.5 beta. It turns out to be easy to use `merge` to between 3.6Alpha1 and the changes in current tar file on sgi.com to generate 3.5Beta source. I have not yet tested the result. vjs