crash_sender: upload oldest crashes first
The current logic is sorting crashes by date (which is good), except it
sorts it such that the newest is first (which is bad).
This means if a new process continues to crash, old crash reports will
be perpetually stalled (it's a stack). Instead, let's start with the
oldest first (like a fifo).
TEST=`cbuildbot x86-generic-full` passes VMTests
Reviewed-by: Ben Chan <firstname.lastname@example.org>
Commit-Queue: Mike Frysinger <email@example.com>
Tested-by: Mike Frysinger <firstname.lastname@example.org>
1 file changed