A world on FFI about raptorjit HOT 6 OPEN

hi! :)

what about that ffi stuffs cant be redefined/removed/gc'ed (dunno, but something like :D ) and there is a limit for definitions, and/or this could lead to name collisions also. (ive just read something like this, kinda much back in time on the lj mailing list.)

is that possible to make these smoother? or even use namespaces, gc, manual removing and whatever? ive got no much idea how it works, but i believe that such things are possible, just not implemented, cuz the lack of actual need or cuz performance, but mayb there are technical caveats behind this...

thx for any info! :)

from raptorjit.

I wonder if we could use Git to curate a collection of compatible (potentially patched) libraries?

Maybe we could reuse @capr's luapower? Or take inspiration from it at the very least.

Snabb is a bit extreme. For each of our dependencies we import the library and its complete git history using git subtree. This is powerful for two main reasons. First if we find a bug in a library we fix it immediately in our own copy (no workarounds) and we can then asynchronously push that fix upstream. Second when we use tools like git bisect they can see every commit on every dependency to pin-point problems that appear after updates.

I'm sure not every project wants to operate that way al the time but I also think that dependency management in the Git era isn't necessarily like it used to be.

from raptorjit.

I think the luarocks top download can be also a list of targeted module/feature.

from raptorjit.

I wrote a luasocket ish wrapper around unix sockets for unix and windows. A bit untested though

https://github.com/CapsAdmin/luajitsocket

from raptorjit.

My current goal is just listing existing FFI solution equivalent to usual modules.
I'm open to discuss about FFI but it was not my original goal.

@CapsAdmin that is the kind of info/project I search.

If RaptorJIT drop the Lua C-API, I hope to have an index of existing solution for common needs.

I would like to avoid the following way :

• I need some missing feature
• I search to existing solution
• I found plenty of them
• I'm lazy or don't have time to test them
• Often testing all existing solutions spend more time than writing one from scratch
• I wrote my own, it is better for me
• There is another alternative now but it is maybe not better for the world. https://xkcd.com/927/

from raptorjit.

I think a better view of usual needs is the dependencies top : https://luarocks.org/stats/dependencies

from raptorjit.