Software crashes are feared by software organisations and end users. Many software organisations have embedded automatic crash reporting tools in their software systems to help development teams track and fix crash-related bugs. Previous techniques, which focus on the triaging of crash-types and crash-related bugs, can help software practitioners increase their debugging efficiency on crashes. But, these techniques can only be applied after the crashes occurred and already affected a large population of users. To help software organisations detect and address crash-prone code early, we conduct a case study of commits that would lead to crashes, called "crash-inducing commits", in Mozilla Firefox. We found that crash-inducing commits are often submitted by developers with less experience. Developers perform more addition and deletion of lines of code in crash-inducing commits. We built predictive models to help software practitioners detect and fix crash-prone bugs early on. Our predictive models achieve a precision of 61.4% and a recall of 95.0%. Software organisations can use our proposed predictive models to track and fix crash-prone commits early on before they negatively impact users; increasing bug fixing efficiency and user-perceived quality.