If the following has been discussed before please just remind me of the conclusion.
I'm just wondering what the philosophy is, or should be, when a fatal(?) error is encountered in a library. In libsec I noticed a few calls to exits (mostly when out of memory). In a threaded program this might blow away one thread while the remainder of the program is left (limping) around. Wouldn't sysfatal be better? (blow away whole program) Or never exit from (in) a library, but always return? Just curious, Axel.
