Description

thbourlove

opened on Nov 19, 2014

Issue body actions

No description provided.

Activity

1. igorw commented on Nov 19, 2014

   

   igorw

   on Nov 19, 2014

   Contributor

   More actions

   First of all, thanks so much for reporting this!

   This highlights a pretty interesting flaw in the current design of the stack. The stack assumes that every middleware will handle its own layer of termination. The stack just triggers every terminate() individually. However, this is not the case. Since a middleware is a decorator, if it implements terminate(), it will of course also call terminate() on the next layer!

   As a result, it will be called once by stack, and once by the cache middleware itself.

   The reason for doing this in the first place, is that terminable is optional. So because not every middleware of the chain has to implement it, the chain can be broken. The point of the stack is to never break this chain, and call all of the terminable middlewares.

   I have some ideas for fixing the problem, but it is going to get a bit complicated.
2. igorw commented on Nov 19, 2014

   

   igorw

   on Nov 19, 2014

   Contributor

   More actions

   It's... complicated.
3. igorw

   added 2 commits that reference this issue on Nov 19, 2014

   Build safe chain of terminable kernels (fixes #10)

   This issue will close once commit acea575 is merged into the 'master' branch.

   acea575

   Assume Terminable middlewares will delegate (fixes #10)

   This issue will close once commit 17887ec is merged into the 'master' branch.

   17887ec
4. igorw

   closed this as completedin 17887econ Nov 23, 2014
5. igorw

   added a commit that references this issue on Nov 23, 2014

   Update CHANGELOG for v1.0.3, refs #10

   c1f8a46
6. igorw commented on Nov 23, 2014

   

   igorw

   on Nov 23, 2014

   Contributor

   More actions

   This issue has been fixed in v1.0.3, thanks so much for reporting it!

Issue actions