Archived from groups: microsoft.public.win98.gen_discussion (
More info?)
On Sat, 26 Mar 2005 19:47:17 -0500, meirman <meirman@invalid.com>
>on Fri, 25 Mar 2005 22:07:50 "cquirke (MVP Windows shell/user)"
>>On Thu, 24 Mar 2005 21:45:25 -0500, meirman <meirman@invalid.com>
>>>I've temporarily gotten past my "Registry Checker" problem.
>>>Now the problem is that booting stalls further on. Bootlog.txt ends
>>>at "initing esdi_506.pdr". Does anyone know why?
>>Most likely there's a common source of both problems, and it's at a
>>deeper level of abstraction - such as flakiness at the hardware level.
>>Such problems can render any writes to the HD unsafe
>>So I'd step back and do "the prelim" to check that the abstraction
>>layer you are diggining in is actually resting on solid ground:
>>http://cquirke.mvps.org/9x/bthink.htm
>Thanks for this page. I've been reading it, and the pages it links to,
>The one on the Startup Axis will take some time to complete.
The crucials in your case are the hardware tests, then the formal
malware scan
http://cquirke.mvps.org/9x/virtest.htm
>The short version is that I ran scandisk, it still crashed the next
>time, I started in Safe mode, ran msconfig, checked "force
>Compatibility mode disk access", and this time it started normally. I
>unchecked it, and it still starts ok.
>Does this make sense to you?
Yes; insane for a static cause of failure, implying it's a variable
problem. Not easy, but not uncommon
😛
>During startup the last time, it displayed one Unkonwn Device, with a
>big yellow question mark, and started the Install Hardware process,
>but whereever I look, I can't find the files it wants. I have no clue
>what device it is talking about, but it said it wouldn't ask anymore.
Now that interests me! By their resources shall ye know them, e.g. an
unknown device that uses IO port 220h, IRQ 5 and DMA 1 is most likely
an unrecognised or flaky sound card, etc.
Does your bonged (!) device use any hardware resources?
Is it there in Device Manager every time, or variable?
If variable, I'd think flaky hardware, bad NVRAM storage, or both.
>And normal Windows still started. I've restored system.ini to the
>boot process and it starts fine, and I plan to put back everything
>else, one step at a time.
OK.. With System.ini out of the frame, you'd have had 16-color vanilla
VGA. But I'd stay out of Windows until I was COAB certain about the
hardware. What hardware spec is this PC? Pentium III generation? If
so, then that's new enough for this issue to apply:
http://cquirke.mvps.org/badcaps.htm
>I got around registry checker by disabling it in the scanreg.ini file,
>but I suppose I'll have to reinstate that too, right?
Hmmm... when all is well, maybe. Spontaneous registry corruption
makes one VERY worried about bad RAM and similar low-level flakiness;
what happens is the registry image is seen as bad due to bad RAM or
similar effects, so in response the OS splats the one on HD, which may
or may not have been bad.
A structurally-bent registry can initiate problems, a la...
http://cquirke.mvps.org/9x/registry.htm
....but hardware must be OK before tackling that.
>In trying to follow your recommendations, I had tested the memory a
>couple days ago for 10 minutes and then for 3 1/2 hours, but I dl'd
>newer versions of Memtest and DocMem. I had trouble using them, but
>I'm not asking about that now.
OK. I'd be happier with an overnight run of all tests in MemTest,
whichever takes longer. Both have to be run from boot disk, which
they make themselves. If you have trouble there, do ask
>I also ran scandisk, in DOS and then in Windows Safe Mode. I let it
>fix a few small things -- having read your cautions about not letting
>it fix certain things -- and none of them involved files that sounded
>remotely important. For example a .url file in the Real directory,
>some files which it said didn't have proper names, 8.3 names but one
>was a .dll, other little stuff. I have the log file
It would be of interest (er... trimmed it pasted) for two reasons:
1) Pattern of errors
Some errors can arise from the interruption of normal file operations,
i.e. bad exits - but others indicate something more is going on. This
is covered in some detail in the "FAT data recovery" section of the
site. Which files are involved, may also point to what was in
progress at the time a bad exit occured.
2) Secondary damage
You may get a static failure pattern (i.e. the same context always
errors in the same way) if a significant file is bent.
>Scandisk/windows restarted itself several times,
I don't do Scandisk in Windows. By definition, if I'm doing Scandisk,
it's because I consider the file system or HD to be at risk, and
therefore unsafe for writes, and therefore unsafe for Windows.
>...continued with the surface scan where it had left off.
Windows Scandisk surface scan (called "thourough test" there) is not
only dangerous, but gives you far less info:
- course-grained progress counter can't reveal retry latency
- timing will be affected by background tasks anyway
- no map showing any existing bad clusters
>It found no bad sectors.
Did the final report state no bad clusters as well?
>Scandisk/do had found no bad sectors also, but I turned
>it off about 80 or 90% of the way through, when it was in unused
>space, since I knew I would run it again in Windows.
Ergh, but OK. Bad clusters in free space beyond the data "edge" will
prolly not affect you unless defragging (don't defrag!) until the data
edge sweeps forwards to use them.
>After running scandisk, I tried to restart, in step-by-step, and it
>crashed in the same spot on the screen and the same spot in the
>bootlog.
Ahhh.... details there?
>So I started in Safe Mode and ran msconfig looking for something,
>anything I could disable that might help it start. I chose to check
>"Force compatibility mode", on the theory that esdi_506.pdr is related
>to something better than compatibility mode, and if I force compat, it
>might not try to run esdi506.pdr.
Good call! If you've never allowed Scandisk to overwrite
C:\Scandisk.log, you could check that to see if previous passes had
ever "fixed" that file, esp. for crosslinks. What's the size of the
file? Is that size correct for your Win9x version, or smaller? If
smaller, is it a round multiple of the cluster size? If so, then
expect truncation damage from a Scandisk "fix".
>Voila, it started, all the way to Normal Windows. Along the way, and
>again the next time, it tried to install one more piece of hardware,
>which it called Unknown Device, and displayed a big yellow question
>mark. I tried the win98 CD and every folder in windows that it showed
>had any .inf files, but none worked.
Certainly sounds like all is not well somewhere - bad or mis-driven
UIDE controllers being the obvious guess. VIA chipset? New
motherboard dropped in as replacement with the same Win9x
installation? I'd wonder about bus speeds (overclocking) too. Did we
cover hardware specs yet? We should, around about now
🙂
>---------- ----- ---- --- -- - - - -
Gone to bloggery:
http://cquirke.blogspot.com
>---------- ----- ---- --- -- - - - -