home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.os2.misc
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!sgiblab!munnari.oz.au!metro!usage!spectrum!pendrith
- From: pendrith@spectrum.cs.unsw.oz.au (Mark Pendrith)
- Subject: Kernel Roulette
- Message-ID: <1992Nov16.181114.7476@usage.csd.unsw.OZ.AU>
- Sender: news@usage.csd.unsw.OZ.AU
- Nntp-Posting-Host: henna.spectrum.cs.unsw.oz.au
- Reply-To: pendrith@spectrum.cs.unsw.oz.au (Mark Pendrith)
- Organization: unsw
- Date: Mon, 16 Nov 1992 18:11:14 GMT
- Lines: 96
-
-
- Since first running OS/2 2.0, I have had four diffrent kernels installed.
- They are:
-
-
- GA (internal rev. 6.307 March 1992)
-
- PJ03686.zoo (internal rev. 6.311 May 1992)
-
- OS2KRNL.zoo (internal rev. 6.412 June 1992)
-
- "Golden" CSD (internal rev. 6.6?? (from memory) October 1992)
-
-
- All these kernels have their problems, some more than others. They are all
- broken. By "broken" I don't mean that they have quirks or documented minor
- problems with documented work arounds; I mean big-red-switch-reboot type broken.
-
- The worrying trend is that the later ones seems to be *less* stable than the
- earlier ones.
-
- HISTORY
- -------
-
- The original GA kernel's problems are well-documented, so I won't cover all of
- these here again. The most serious and visible problem in GA was the famous
- non-shrinking SWAPPER.DAT file.
-
- The May fix (6.311) was provided to fix this problem, but other known problems
- in GA were not addressed until the June fix (6.412).
-
- Unfortunately, the June fix introduced yet another pathological bug, the
- dreaded PMV2001 error.
-
- PMV2001 popped up reasonably regularly to inform you that you were out of
- available memory (on a VM os with 8MB RAM and 40MB free disk space - amazing!).
- PMV2001 suggests you close some of the currently running applications to free up
- some memory. Which you would do, but to no avail. You close application after
- application, spurred on by PMV2001, until PMV2001 tells you that simply running
- the WPS is taking up too much memory. So you shut down and reboot. Only now is
- PMV2001 satisfied. Until, that is, the next time you decide to run, say, two
- windowed DOS sessions in the background while attempting to open a W4W 2.0
- session under WIN-OS2 in the foreground...
-
- (The guys who devised this particular error message must have been inspired
- by Kubrick's film of the same name. "Open the WIN-OS2 session, please HAL...")
-
- The October fix, the much anticipated "Golden" CSD, its very name dripping
- with Homeric promise, was supposed to deliver us from all this evil. Well,
- the gods hate hubris, and, boy, were the guys who wrote this punished for
- their presumption. The "Golden Turkey" CSD has many quirks and surprises for
- the unwitting, but the most worrying in my experience is the development of
- the the SYS3176 error: A program executed an illegal instruction at xxxxxxx.
-
- SYS3176 is trying to tell you that an illegal opcode was about to be executed.
- HELP 3176 advises you to inform the software developers that their code has
- illegal opcodes where legal ones should be. Strange, but these same programs
- did not contain illegal opcodes when run under DOS or the previous three
- kernels. Of course, there are no illegal opcodes.
-
- What I suspect is happening is far more worrying. It seems far more likely
- that the task switching algorithm is somehow getting its table of CS:IP
- pointers screwed up from time to time, and is therefore randomly attempting
- to execute bytes in data segments or even unallocated memory as code. This
- is speculation, of course; if anyone has a better theory, I'd be glad to hear
- it. But its all academic. The bottom line is that the "Golden" CSD is unusably
- flaky (for my purposes, at least) and the worrying trend seems to be that at
- least as many severe bugs are being introduced in each fix as are being
- eliminated. And so I'm back using version 6.311. It's not much better than the
- first version, but it's a lot better than the most recent.
-
- Is it too much to ask for a reliable kernel after having to suffer these four
- attempts? I'm sure I can't be alone when I say my confidence in the medium-term
- viability of OS/2 has diminished as a result of this recent "Golden" CSD farce.
- I believe IBM is losing both time and prestige and is in danger of squandering
- an ever decreasing window of opportunity with this sort of blundering. It's
- strange that everyone is asking "Can Microsoft kill off OS/2?" The irony is
- that only IBM can kill OS/2; and continuing to release unstable version after
- unstable version is the way to do it. Let's hope it doesn't happen.
-
- On the upside, I can happily report that the telephone support from IBM
- (in Australia at least) has been exceptional. This is the way software support
- was meant to be. I have never had a call back from their support centre take
- more than 24 hours, and that's when I've said it wasn't an urgent problem.
- You get published bug lists, so at least you don't have to scratch your head
- and wonder. The honesty is refreshing. These guys are professional. The only
- problem is, it spoils you forever for dealing with the likes of Borland and
- Microsoft. OS/2 must prevail. An NT future is too horrible to contemplate.
-
- ------------------------------------------------------------------------------
-
- Windows NT : (The NT really stands for "Not Tested")
-
- pendrith@cs.unsw.oz.au
-
-
-