| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 946.1 | not supported | EDSDS6::GLEASON | Daryl Gleason, DECset Engineering | Thu Mar 20 1997 10:27 | 16 |
| Hi Andre,
Note 615.1 in this conference has some summary information that may be
helpful, but I don't think there is any documentation on exactly how
PCA behaves apart from the statement that collection of data for kernel
mode code (other than page fault data) is not supported.
Exactly what happens when such collection is attempted depends upon
what data you're trying to collect. For example, if you're doing
sampling, which is based upon a 10-millisecond timer, the timer ASTs
cannot be delivered when kernel mode code is executing, so any pending
sampling ASTs will be delivered all at once when the code returns to
user mode (which obviously can skew the results if the code takes
longer than 10ms to execute).
-- Daryl
|
| 946.2 | What about coverage. | INTONE::FONTAINE | | Thu Mar 20 1997 17:07 | 5 |
| What we are interested in is test coverage only. We're
curious about what happens in kernal mode when measuring
coverage.
-Andre
|
| 946.3 | | EDSDS6::GLEASON | Daryl Gleason, DECset Engineering | Fri Mar 21 1997 08:28 | 10 |
| In order to measure coverage, PCA needs to put a breakpoint in the code
to be covered and subsequently trap it so that PCA can detect that the
code has in fact been covered. Since PCA cannot trap breakpoints in
kernel mode code, coverage data cannot be collected for that code.
I'm not sure what the exact behavior might be if you try to do this;
it's probably unpredictable. At any rate, I recommend that you avoid
trying to collect coverage data on kernel mode code.
-- Daryl
|
| 946.4 | Take alook at note #912.10 | SMURF::FENSTER | Yaacov Fenster - System Engineering, Troubleshooting and other m | Thu May 22 1997 19:18 | 5 |
| > <<< Note 946.0 by INTONE::FONTAINE >>>
> -< Kernel questions >-
Take a look at note 912.10
|