It may be of little help, but the hex equivalent of "-1073741819" is 0xC0000005 which means "Access Violation" Exception and is typically associated with a program trying to read or write to an invalid memory location (I'm not coder so others may be able to give more details)
we now have eanabled Debug Logging for the bmi and can indeed see, that a "not well-formed" read INCOMPLETE XML is handed over to the server.
Right now i believe we have hit some sort of length limitation.
We found a working System-Package.
Doing the same debugging on that also shows an incomplete XML, but as this one is working it may be an issue with Debug-Logging of bmi.
Unfortunately we still have no clue, what stops us at the moment.
Allow me to understand....Please verify the following for me:
You are doing a bare-metal install of windows server 2008r2 (or is it some other version?)
You are using a winpe image you have generated (assuming you use WAIK v3 or v4)?
You are dealing with two system packages, both of which generate the invalid token error (line 532), yet one of them proceeds and the provision complete, while the other causes the job to fail.... and this is repeatable? In other words the job that worked, always works?
Is this correct?
Are you manually editing the unattend.xml in the BSA gui?
Is the target host to be provisioned a virtualized host? If so, what hypervisor?
Are there any discernable differences between the system package that does work and the one that does not? (specifically, does one have a license key while the other does not?)
the bmi in your image matches the bsa appserver version ? (did you regenerate the winpe images w/ the new bmi after upgrading ?)
Yep that's all good....
I fixed the Issue now myself.
Was a bit tricky.... will post details tomorrow.
troubleshooting this issue has helped me a lot in understanding BMI a bit better now.
So when i've set the LogLevel to Debug, i saw that already on the very first step (Switchboot), the bmi process seems to load ALL instructions for the installation from the App-Server.
With ALL i mean:
- Pre-Install Scripts
- Unattended XML
- Post-Install Script
- and all the magic bits and pieces the BL adds on it's own between the various steps.
The whole instruction set is handed over in one BIG XML block.
The reason why i thought we have hit some length issue, was due to the fact that the bmi.log displays that whole XML block in a single line, with up to 30.718 colums.
But that line ended like this:
<data>set log=runonce.log</data> <data>set counter=0</data> <data>:errnet</data> <data>if
However, this seems to be some sort of issue, with the Debug-Logging, as also on a working System-Package this line ends abruptly.
Looking at the content in more detail it even stopped logging before the Post-Install Script Content.
But what i found described below, is that the issue actually was in the Post-Install Script and that this issue has prevented BMI, from even finishing the very first step successfully.
After i had found a working 2012 System-Package in the System i dumped both the working and the broken ones as XML and ran a compare between those in Notepad++.
And this has brought to light the following line:
There was a dodgy character in the System-Packages Post-Install script that had been added by our Windows-Engineering.
In the BSA Console the entry looked perfectly alright, but probably due to some Code-Page mix-up it was not.
After correcting this entry, everything ran smooth again.