brrt good * 06:17
nine MasterDuke: the description reads as if it works similar to NativeCall in MoarVM before I added JIT compilation capability. So we're already nicely ahead there. What could be interesting is the part that gathers function signatures from debug information. 06:38
nwc10 has not read it yet, but I've been thinking about the possibility of that for some years
and I think mentioned it to timotimo at GPW2018
only as "it ought to be possible" 06:39
brrt what is or isn't possible? 06:40
nwc10 getting C function signatures direct from the debug section of a shared object (if one is present) 06:40
so avoid the DRY violation of having to also write them in the Perl code
or having to find the relevant C header file and parse it
brrt that is an awesome idea right until you install it on a server and the administrators / policy has decided not to ship header files and debug sections 06:42
nwc10 which is a very logical thing to do on a production server 06:44
nine So it would make more sense to gather that information at build time (probably from header files), just like in C. 06:45
Now this core feature list reads darn impressive: github.com/pybind/pybind11
brrt on the other hand, if you could precompile the nativecall'bindingness on a dev server (where you'd have headers), and ship that as-is to the production server, then, that's probably a good idea 07:12
x64 ABI support aint that hard though 07:13
(windows x64 ABI)
it is arguably simpler /better than the unix one
nine brrt: that's what I meant with "gather that information at build time" 07:14
brrt it is a good idea 07:16
nine Btw. we already have something in that regard: github.com/Skarsnik/gptrixie 07:25
Geth MoarVM: samcv++ created pull request #841:
Randomly seed a hash secret to mitigate ease of DOS attack
07:55
AlexDaniel can somebody look at this please? github.com/rakudo/rakudo/issues/16...-379541036 11:20
robertle regarding 1663, I can help with debugging/testing. just holler if you don't have a mips or s390x around 13:32