We really need to replace these old non-generic collections with generic ones.
BlueRow's IDictionaryEnumerator implementation iterates over values when passed
to a foreach loop, not key value pairs.
One of the 'securities_incremental' fns was calling back into itself instead of
the other overload. This is a 'preemptive' fix because the problematic code path
wasn't running in practice (though it was a ticking time bomb).
* Add BlueRowReflection<T> so ViewStream implementations can use
strong row types without needing to implement IBlueRow
* Add default implementations for IDictionary members in IBlueRow to make
implementing IBlueRow less tedious
* Get rid of old 'cfg_defaults' param in get_proto_sources impl for linux client
builds
* Make libfmt dep header-only
* Use CMAKE_SHARED_LIBRARY_PREFIX if "PREFIX" is undefined when copying libspark
to dotnet bin dirs
This (hopefully) fixes part of a compounding error where 1) the SprayAndSkewAlgo
sends 0 lot orders, which are rejected by the exchange. Then 2) the REJECT exe
report fails a misguided "validation" check due to its 0 qty, and is never
propagated to the rest of the system.
'PostOnly' is a TimeInForce flag that we don't have in our system. It makes it
so Limit orders that would be filled on entry are rejected. Ali's SprayAndSkew
algo needs this to avoid taker fees.
This particular error log was intended for failed position requests. Since an
empty list evaluates to false in python, we were incorrectly interpreting an
empty result as a failed request.
* Change 'new_req' -> 'req'
* Change 'self.exe_report_callback' -> 'self.bridge.on_execution_report'
...this is what you get when you take shortcuts and copy/paste code.
* Round fill price and qty to 7 digits in bybit gw plugin
* Remove erroneous line of code that reset filled_quantity to 0 on each
update (it was causing partial fills to be overcounted)
I found myself constantly changing it from WARN to INFO while developing/testing.
This is a 'dev' config profile, so it should be tuned for development.