fred Async chkinsertsender riskier cleanup

This branch makes CHKInsertSender and CHKInsertHandler largely asynchronous, and greatly reduces the number of threads needed on high bandwidth nodes. It has been tested reasonably thoroughly, both on my local node and in the synthetic RealNodeULPR/RequestInsert tests. Please seriously consider this for a test build for wider testing, and let me know about cleanups/style issues I've missed. Thanks!

There are two other async-chk* branches that might be marginally lower risk.

Aug 11 '15 20:08 toad

This seems to work on my node and also in synthetic tests (RealNodeRequestInsertTest and RealNodeULPRTest). Please consider for merging into a test release. It should greatly reduce the number of threads needed on high bandwidth nodes.

Aug 15 '15 12:08 toad

Unfortunately this might be affected by issues in #400. One possibility is to leave out the last commit but I'm concerned about memory usage. So this might be blocked... :(

Aug 16 '15 18:08 toad

@toad Bombe said that #400 is OK now, is there something else blocking this?

Nov 10 '15 13:11 ArneBab

Not IIRC, although it should be tested carefully, especially in vitro.

Nov 16 '15 15:11 toad

Tested for days without problems in vivo. Number of running threads was reduced drastically by this changeset — well done.

I did not review the code except for quickly reading through, nor did I test anything in vitro except for unit tests.

Dec 05 '15 23:12 bertm

Merged next into the branch. Please consider for 1472.

Dec 30 '15 23:12 toad

Whee, more conflicts.

Jul 18 '16 04:07 Thynix