RetryException is not serializable bc Attempt is not

[eschultink]

RetryException implements Serialzable, but does not fulfill the contract, because it has a field Attempt<>:

re-retrying/src/main/java/com/github/rholder/retry/RetryException.java

Lines 32 to 34 in 5c75a68

	public final class RetryException extends Exception {
	private final Attempt<?> lastFailedAttempt;

which does not implement Serializable:

re-retrying/src/main/java/com/github/rholder/retry/Attempt.java

Lines 30 to 38 in 5c75a68

	public class Attempt<T> {
	private final T result;
	private final Throwable throwable;
	private final int attemptNumber;
	private final long delaySinceFirstAttempt;

as well as being generic, which makes it non-obvious how to solve this. A few options:

1. just make Attempt<> implement Serializable, ignoring potential problem with T not also being Serializable. Then it's up to users - if they're using Retryer with serializable return types, things will work; if not, no worse off than now.

2. make result transient. should work, with caveat that if the instance has undergone serialization and deserialization, result will be lost, so in some sense hasResult()/getResult() won't respect contracts, as will be false/throw IllegalStateException, even though original call did have a result. But again, this only happens after Attempt has undergone a serialization+deserialization cycle, which currently throws an error, so I would argue that everyone's still better off.

3. require T to implement Serializable. Probably not great solution, as would constrain users too much and breaks Retryer's interface.

I'd personally vote for (1) as the simplest, lowest impact change; but quick version of (2) might be just as good, if later followed by bigger change to make hasResult()/getResult() behave sensibly.

[jgiardin]

@eschultink I am experiencing this serialization issue as well - Did you ever get resolution or a workaround? Or perhaps recommend a different retry library?

Thanks!