Why do you assert! if the thread is panicking in your `Drop` implementation?

[jymchng]

https://github.com/stouset/secrets/blob/master/src/secret.rs#L215

For the Drop implementation of Secret<T> (linked above), why do you assert if the thread is panicking when the target is not on Windows and munlock was successful?

My purpose is to understand and learn from you.

Thank you.

[U007D]

The comments say that attempting to munlock() a second item on the same page on Windows is expected to return an error, and is being ignored.

As I read it, the code says that if munlock() failed and this isn't on a Windows system (and the last OS error code was 158), then that's a problem (ie. we must panic) unless we're already panicking.

[stouset]

To add to the answer here, until #110 it was possible for two consecutive secrets to exist in the same memory page. When one was dropped, it would munlock the page that both were in. The second munlock would fail, but only on Windows, where the function to do so produces an error.

This check looked for just this error and swallowed it.

With #110 we align secrets to page boundaries, so two secrets cannot occupy the same page. So this check is no longer needed.