[addrmgr] Unreachable control paths in addrmgr.updateAddress

[DanielKrawisz]

Say you try to update the same address multiple times with addrmgr.updateAddress. Well in lines 198 - 201 in addrmanager.go there is a condition for the function to return for a certain probability without adding the input address in a new bucket. If the function does not return at that point, we go down to line 213 where the getNewBucket function is called. Then, in lines 216 to 218, the function returns if the given bucket already has that address in it. The problem is that getNewBucket always returns the same number for a given input. Therefore, if you try to add the same address twice, the function always returns without adding the address in a second bucket because getNewBucket always gives it a bucket that already the address in it.

This seems to be a bug because the function would not be written this way if getNewBucket was always expected to return the same number always for any given address. Also, there is a constant called newBucketsPerAddress which is set to four, so clearly it is expected that there could be up to 4 new buckets that could have this address in it. Also, in the Good function, we go through every new bucket to check if a given address is in it, so it seems as if potentially the address might be found in any new bucket. However, there is no way for the new address to get in any bucket but the one whose index is always returned by getNewBucket.

getNewBucket does not have comments explaining what it is supposed to do, so I'm not really sure what the expected behavior is supposed to be.

[davecgh]

@dajohi: Can you look into this please.

[dajohi]

I am currently looking for this subpkg to be completely rewritten.