Anthony Pesch

Finished the work to merge the lvalue decoding logic between pb_decode.c and pb_decode_fast.h. The size of the pb_decode.o compiled by the unit tests is a ~4 kb larger due to...

Will do. Do you have any feelings regarding the file names / preprocessor defines? I went with PB_DECODE_FAST at first, and then I tacked on PB_DYNAMIC_DECODE_FAST but I didn't put...

Chopped up the commits into the relevant changes I believe.

@PetteriAimonen thanks a lot! Please let me know if there's any follow up tasks you'd like to see done, etc.

Started pushing a few wip experiments to that branch that aren't required, but I thought I'd point out that https://github.com/nanopb/nanopb/pull/1008/commits/515914eff6f16bf7a12fac4d1e8d033736f749e1 looks like a legit bug in both code paths. If...

Ripped out my WIP changes to a new branch, and pushed both that fix for pb_release and another change to directly call a submessage's decode handler.

@PetteriAimonen ripped out the memcpy and replaced with an if conditional. One thing I didn't understand when I was reverting that code - why did the previous code cast to...

@PetteriAimonen ah, correct, data_size is compared against that. So forgive my ignorance, do these platforms not have *int16_t and *int8_t defined then? Edit: Pushed with the code again using the...

Sorry, was this supposed to say in big-endian mode (instead of in little-endian mode)? For big-endian mode, it seems changing https://github.com/inolen/bit-buffer/blob/master/bit-buffer.js#L166 to: ``` getFloat64 (offset) { BitView.#scratchU32[!this.#bigEndian] = this.getUint32(offset); BitView.#scratchU32[this.#bigEndian]...

What machine are you running this on? It feels like you're running this on a BE host.