Comment 1 for bug 1235436

So just to confirm, after manually touching /etc/apport/autoreport, my phone is now being crushed by several instances of the apport-noui job processing *the same crash file*, in parallel - apparently, for some reason whoopsie-upload-all finds it necessary to run gdb over the crash files, which is not pretty.

This reinforces my belief that we should ditch /etc/init/apport-noui.conf completely, and let whoopsie handle those crash files directly.

But it also looks to me like /usr/share/apport/apport does not write its crash files atomically... it opens the .crash file, writes to it, and closes it. Which means that regardless of who's watching /var/crash with inotify, there's a race condition here that means some things may try to process the crash files before they're fully written. And on armhf, writing these crash files out is sometimes VERY slow. I think we need apport fixed to write to a temp file and atomically rename into place as .crash, to ensure this all works reliably.