Wisdom has it that rare random crashes at clients’ sites can be debugged
by having them save a crash dump file, then analyzing it. However, what
is the recourse when the crash dump file itself cannot be saved?
Case in point is that I am writing an all user-mode application. At one
site the client encounters occasional IN_PAGE_ERROR exceptions (which I
highly doubt are related to my application, only that’s the one they run
all day long). Attempting to save a crash dump file with Sysinternals’
ProcDump or Windows’ own Task Manager results in “Error writing dump
file: The system cannot read from the specified device”, while ADPlus
logs an attempt to save the dump file but silently fails to do it.
While I could not duplicate their specific error in house, I can force
an IN_PAGE_ERROR on my local development machine as follows…
- copy all of app’s files to a memory stick;
- run the app off the stick;
- unplug the stick from the USB port;
- next action which requires paging in parts of the not-yet-loaded code
or data would trigger an IN_PAGE_ERROR;
…and sure enough, attempting to save a crash dump file at that point
fails with the same quoted “Error writing dump file: The system cannot
read from the specified device” message. What would then be the
alternative to gather at least some context information, or (ideally)
a clue as to “page missing, on device X referenced by driver Y”?
I have posted the same question on Sysinternals’ forum and the Microsoft
new-fangled “newsgroups”, without much echo. Any advice appreciated.
Thanks,
Liviu