ken@gvax.cs.cornell.edu (Ken Birman) (11/30/89)
In connection with the forthcoming MACH release from Mt Xinu I decided to make a slightly updated ISISV1.3 release. Call it ISISV1.3.1. This includes the pmk patches that have been posted (pmk is an ISIS demo), the sun_grid demo correction, a new UUCP file copy facility for the ISIS long-haul and spooler program, corrected man pages for a few entries that were out of date, and the MIPS stack alignment bug fix. If you are running ISIS V1.3 I do not see any reason to switch to 1.3.1 These fixes have all been posted as patches and not one of them is important (well, the MIPS one is, but none of the others). The TAR tape will still be called ISISV13.TAR.Z, however, because ISISV131.TAR.Z looks unacceptably ugly. It is obvious which release you have because this one creates top-level source directories called isisv1.3.1 and meta1.2 instead of v1.3 and 1.0 respectively. Please note that we may have introduced bugs when creating this TAR image. If you pick it up and notice anything wierd at all, please let us know immediately. ---- Note to ISISV1.3 users: Nobody has reported a memory leak problem other than the MIPS problem corrected by the stack alignment patch. Hopefully, this takes care of that. Here is the entire list of known problems: 1) isis -A dumps core on SUN systems when compiled with -DNOSUNLWP. I am going to post a patch shortly. This may also affect MIPS and HP users It only dumps core when "protos" actually crashes, though. The person who noted this was provoking a lot of protos shutdowns and hopefully the problem isn't noticable in normal settings. 2) On the most recent release of HPUX, it seems that ISIS has become buggy, with a variety of core dumps coming from the "startup" program bin/isis. This may be related to a new HPUX facility called BIND. Typical problems are that the "cmd" tool dumps core when you say "list @*", which works find on all machines at Cornell, and after an isis "shutdown" you tend to get a core dump from cmd and/or bin/isis. I believe ISIS V1.3 is quite solid on the pre-bind versions of HPUX. 3) There may be situations in which isis -Z doesn't actually shut down the system, but I am having problems reproducing this. The report was from an HP user, so it could be related to (2). 4) There may be situations in which isis_accept_events(ISIS_BLOCK) blocks forever. I am testing this now and will patch a bug fix if necessary. These are all pretty minor looking bugs. If you know of anything at all that I haven't listed, you may be the only person who knows of your problem! Please, email to isis-bugs if you have something to report! Otherwise, we will continue under the illusion that ISIS V1.3 is a remarkably solid version of the system. Ken