parsedump: Simplify by using poi instead of regex

[EchterAgo]

Instead of using dt on OpenZFS!cbuf and checking the output we can just use poi(OpenZFS!cbuf) in the .writemem command.

[EchterAgo]

I also added this to https://openzfsonosx.org/wiki/Windows_BSOD#Debug_Print_Buffer

[andrewc12]

Dang thats better.

Does this script work for you?
I kept having problems with paths/names being wrong for python and cdb.

[EchterAgo]

Dang thats better.

Does this script work for you? I kept having problems with paths/names being wrong for python and cdb.

It works for me with some adjustments. I guess we should improve it to use os.environ["ProgramFiles(x86)"], os.environ["windir"] and maybe read the OpenZFS install location from registry. Maybe also have a --zfspath command line override. I can add that.

[andrewc12]

I think cdb installs in a different place if you install it as part of the win sdk vs directly from the msi.

We might need to check multiple locations

[EchterAgo]

I'd also really like it to support a KDB connection to a different machine. That would be really helpful to automate remote analysis on certain conditions like low memory, I'd need that for #283

[EchterAgo]

I think cdb installs in a different place if you install it as part of the win sdk vs directly from the msi.

We might need to check multiple locations

Do you know the other location? I only have these:

[andrewc12]

Yes
"C:\Program Files\Windows Kits\10\Debuggers\x64\cdb.exe"
like in this script

[EchterAgo]

I think stack = run("k ; q") is also not needed, it is part of !analyze -v

[EchterAgo]

We could also add some more info like !stacks and !memusage