WIP: Test for OpenRPC subscription methods on trace/debug modules / whitelist-style pubsub skips #364
Conversation
Date: 2021-04-23 06:43:41-05:00 Signed-off-by: meows <b5c6@protonmail.com>
These modules don't have '_subscribe' methods. Removing these special registrations is not for sure what I want to do (maybe I actually just want to change the mock methods provided) but this is a lego piece. I'm removing it now experimentally. Date: 2021-04-23 06:48:42-05:00 Signed-off-by: meows <b5c6@protonmail.com>
I want to allow debug and trace API modules through. Date: 2021-04-23 06:49:26-05:00 Signed-off-by: meows <b5c6@protonmail.com>
…methods This adds a whitelist-style condition to the PubSub handling for unconventional methods and receivers. See the added and commented methods in ethclient TestRPCDiscover to see the outcome. Date: 2021-04-23 07:43:42-05:00 Signed-off-by: meows <b5c6@protonmail.com>
These are now not required anymore since we allow the reflector to handle them directly (b/c the whitelist-style logic the previous commit(s) add. Date: 2021-04-23 07:45:24-05:00 Signed-off-by: meows <b5c6@protonmail.com>
| "eth_getFilterChanges", | ||
| "eth_getFilterLogs", | ||
| // "eth_getFilterChanges", | ||
| // "eth_getFilterLogs", |
There was a problem hiding this comment.
I've left the removed methods commented for 👀
| // This catches MOST of them (except SubscribeSyncStatus) | ||
| return false | ||
| pkgStr := method.Type.In(0).String() | ||
| if (isPubSub(method.Type) && pkgStr == "*eth.PublicEthereumAPI") || pkgStr == "*filters.PublicFilterAPI" { |
There was a problem hiding this comment.
This is the important part.
I'm using the stringer condition because using reflect.TypeOf(ð.PublicEthereumAPI) will cause an import cycle error.
I think its correct that all the methods from *filters.PublicFilterAPI can be skipped since, AFAIU, these are all "unconventional" methods in service of either eth or debug or trace methods which use subs.
There was a problem hiding this comment.
|
Moving our discussion here as well. |
|
Oh man, well |
|
Let me try something later tonight or tomorrow on this PR as I am working on trace stuff now. I think you opened my eyes one more time and the solution is in the intersection of both PRs. Thanks |
|
Closing since @ziogaschr has got an improved version already going on at #356 , which is this PR's base branch. |
I know it's annoying that I've polluted the diff with the regenerated docs... 🤷
The issue this is trying to solve is that #356's https://github.com/etclabscore/core-geth/pull/356/files#diff-98749d04d0efbddfeb407114a7c48c0309646642e7ff67b807f39ecdc5809924R924 adds
debug_(un|)subscribeandtrace_(un|)subscribewhich AFAIK don't exist; those modules don't use the same weird pattern thateth_subscribedoes.Please feel free to cherry-pick, ignore, modify, or whatever here. Just a sketch of concept.