Re: ie6sp1 : Repair Function Blocked - Anyone know of a fix?

From: Robert Aldwinckle (robald_at_techemail.com)
Date: 07/31/04 

Date: Sat, 31 Jul 2004 19:33:35 -0400

"den" <eponymous@discussions.microsoft.com> wrote in message
news:OIqVk9xdEHA.3148@TK2MSFTNGP10.phx.gbl
...
> a) Prospective Ram Problem in a dependent module.
>
> - Nothing showed under Dr Watson
> - <<Return Code 0cx0000005>> didn't figure as a page fault (it looks very
> similar to the references given to entry points in Dependency Walker, though
> I have run through the various dependencies and been unable to locate it.)
> - Have run extensive tests via a RAM diagnostic software programme -
> including multiple pass burn in tests etc - and no problems showed.

That sounds good. I still think that that 0xc0000005 represents a page
fault though. It may have been trapped so Dr Watson wasn't called.
An example of what the code might have been to have caused such a thing
is: assuming the existence of a pointer and then trying to use it without testing
somehow for its validity. The code fortunately protected itself from crashing
the program completely but hasn't left sufficient diagnostics about the
error/recovery situation to allow you to figure out what needs to be repaired.
That is why we need to trace the flow up to the crash.

Unfortunately I have never used your OS or its Dr Watson so I can only go
by analogy from previous experience with Dr Watson under Windows 3.1. ;o
There IIRC there was a way to increase the sensitivity to page faults
and even report all of them. I think it involved editing an .ini file.
Does the offline help give you any clues about how you could do something
like that? BTW I'm only suggesting Dr Watson because I know that
it is a standard tool. If you have something like Visual Studio you
could use its debugger instead.

There is the still the other alternative that I suggested: using FileMon
to get a list of .dlls accessed when the regsvr32 /i /n digest.dll runs
on both systems. It wouldn't be as reliable a list as a Stack Back Trace
from a dump would be but you would undoubtedly have different
lists from both systems and thus have something new to look at.

Hmm... apparently I have a digest.dll on my system too,
undoubtedly not the same but perhaps similar enough.
Anyway, start FileMon; press Ctrl-L to set the filters.
I used Include: regsvr32*dll Highlight: digest.dll
Then Run... regsvr32 /i /n digest.dll
Stop the trace when you get the prompt that regsvr32
was successful (for the good case) and when you get
whatever the symptom you get for the bad case.

Look after the highlighted lines (i.e. after digest.dll entries)
FWIW I see: rpcss.dll, security.dll, SECUR32.dll after them.
There's more of course but I think that they will be for when the
prompt was being generated.

Even if you don't do the FileMon traces and comparisons
yourself it could be worthwhile to compare the properties
of those three modules on both systems and perhaps
replace them.

Good luck

Robert
---

Relevant Pages
Loading