David McPaul a écrit :
Here is another thought on this. What if we had a text file that the kernel reads on startup that contains a list of commands to automatically execute and write the results to the syslog when the debugger is called. Thoughts?
Maybe a kernel option "print stack crawl after panic()" could be sufficient.But I don't see a way to handle paging... Maybe just change page every minute or so ? Paging is also an issue when booting in verbose mode, since the system stays blocked after showing the first page of log. A boot option to disable paging (or just go to next page without user confirmation) could therefore also be useful.
On a side note : IIRC, Mac OS X writes its panics dumps to NVRAM. This could be a good way to save panic information when crashing... But maybe it is only possible on Macs, I don't know.