VBox.log flushing options

Post by **socratis** » 22. Jul 2016, 16:36

The VBox.log gets updated (I tend to believe) once a certain amount of information in a buffer has reached a specific threshold. Again that's just an educated guess, based on watching several logs "live" as the VM was booting and running. That's fine for normal operation.

There are times although that a more immediate response time is required. These include, but are not limited to:

During VirtualBox betas: that's the time that major features/bugs are introduced. There have been a lot of cases, especially lately (just personal anecdotal experience), where VMs abort, yet the "logger" has not had a chance to update the log with the latest and the greatest, and usually the most important last moments before the crash/abort.
During troubleshooting: the same scenario as above, but this is on a released version and most probably you're trying to troubleshoot something in the guest, not on VirtualBox itself.

I'm proposing to have an "<ExtraDataItem name="VBoxInternal2/VBoxLogMode" value="#"/>" debugging option where "#" is:

Your standard, default current "logger" behavior.
You have a timeout of (let's say) 1 sec, where the "logger" updates the VBox.log.
Immediate, on-the-fly, continuous updates of the VBox.log. Default during the beta cycle.

From the command line (at the beginning I don't think we need a GUI, it's an advanced option) this could work like:

   VBoxManage setextradata "VM name" VBoxInternal2/VBoxLogMode [1|2|3]

or

   VBoxManage setextradata global VBoxInternal2/VBoxLogMode [1|2|3]

where of course the global option overrides the per-VM option per-VM option overrides the global option.

Your thoughts on this? I just wanted your input and refining ideas which I'll collect and file an enhancement ticket. Unless there is already one ticket out there which I missed. Pointing me to it would be welcomed as well.