10

my boss, recently: "Partner company is complaining about seg-faults in our software" [the system is still in testing, not deployment]

me: "tell me what steps to reproduce the error and I'll fix it"

boss: "I'd like to prevent seg faults from happening no matter what"

...I literally can't even at this point

Comments
  • 1
    Give them bus error
  • 0
    @edisonn yes, that's the approach I've decided to take, I just feel uneasy about not exiting on illegal memory access... my gut is that if there's a seg fault you should fix it instead of covering it up
  • 0
    @edisonn haha no I didn't realize that because that is what I was told to do, and I haven't been able to convince my fellow engineers that I'm right (they know that catching SIGSEGV is possible but they didn't really think about the ramifications of doing so)...
  • 0
    @edisonn now that I've got a prototype working, I feel comfortable saying no worries, this is for a soft real-time system that is rather isolated from anything the public will ever interface with... I appreciate the perspective though, it's made me much more comfortable knowing that I'm right to be rather anxious about this whole bit but I believe my approach is fairly conservative (isolating the plugin that experiences a segfault, calling dlclose() and reloading the segment)... I'm also going to keep track of the frequency of these segfaults, and if they happen too close together just shut down the plugin altogether... it's actually been beneficial despite this goofiness because it gave me a chance to do some refactoring to make the system more resilient in general...
  • 0
    @edisonn that's clearly the more sensible thing to do... I'll look into that
Add Comment