Repository navigation
dfuse Deep Mind changes - #8788
Conversation
|
I'll rebase all this after the first wave of comments. |
|
Things I'm thinking we might want to look:
|
255f179 to
40c8bb5
Compare
|
This is ready for a new round of review. |
b064e8d to
2c44f80
Compare
|
This is ready for another review round. |
β¦ature_manager - Those two managers now receives a getter that is able to retrieve a dm_logger instance, if enabled. - Now using a constant for the abi serializer used in deep mind logging statements.
The `generic_ram_trace()` constructor can be used when new RAM calls are being implemented. By using such constructor, it's easy to grep in the source code for all usage of it and properly change the call site to a proper `ram_trace` object.
- Removal of dmlog macro, now using `fc_dlog` directly - Put back deep mind logger in `initialize_logging` - Unbuffered stdout is now performed only when initializing deep mind
d146271 to
f72589f
Compare
|
Updated to now merged deep mind branch on |
|
Anything left to be done/changed here. Would like to get that out in the wild. |
arhag
left a comment
There was a problem hiding this comment.
Please don't force push to the branch. It prevents doing incremental reviews to check if the requested changes were implemented correctly and forces a re-review of the whole PR at each iteration to ensure nothing was missed.
And along those same lines, please allow the reviewer to resolve the conversations they started (feel free to reply "Done" or something similar though) because it makes it easier for the reviewer to keep track of which of their requested changes they had already confirmed were properly addressed.
|
@arhag Yeah indeed, I'm used to rebase my own branches, never realized it breaks this review flow. Will follow that as well as the Those are good comments. |
|
@maoueh: No worries. Also, I forgot to mention: make sure to bring in the latest changes from the |
β¦mind.{hpp,cpp} files and more
|
@arhag Updated PR against your latest comments. |
arhag
left a comment
There was a problem hiding this comment.
I also left a suggestion regarding ram_trace here: #8788 (comment).
And don't forget to address the option description as discussed here: #8788 (comment)
- ram_trace change - fixed deep-mind option description - removed ABI serializer for block - moved EOSIO_DEVELOPER to right location(s)
- Added `maybe_to_variant_with_abi` so that ABI serialization for deep mind purposes is never failing, will now log as a warning the error exception and return the object untransformed. - Fix some comments in the code.
We now use `abi_serializer_max_time_us` directly instead.
Change Description
Discussed last week, here the dfuse Deep Mind patchset rebased on develop.
It contains low-level dfuse instrumentation of the nodeos binary for post-consumption through a console reader.
Consensus Changes
API Changes
Documentation Additions