||[Oct. 22nd, 2003|05:21 pm]
I was stracing a process, making sure something was working, and the last thing on the screen after hitting Control-C:|
write(3, "Blood lube. >.< Bad lube choice ........") = 12323
12323 is a really disturbing return code for write().
2003-10-22 05:41 pm (UTC)
I lost my copy/paste buffer and had to retype it from memory.
You know what I mean.
Yes, I know what you mean. I'll have to remember this when I next need a panic string...
2003-10-22 05:36 pm (UTC)
Gah... human tampon.
I dont get it... :(
Probably some programming thing?
It'd be nifty if you could "undo" a process termination. Before sending the SIGINT signal to a process, the kernel could save the context(s) of the process somewhere in memory. If the physical memory pages used for data weren't reused by another process since you terminated it, the kernel could reload the executable, mapping those pages in, set up the thread contexts, and put the process back into the scheduling list (or whatever).
It'd probably be a bit more complex since it'd have to deal with open file handles and such, and might not work in all cases, but I think it could be done.
There is something called 'pointless checkpointing' which can restart processes from cores. It works with a little support from the operating system to reopen file descriptors, etc.