Mand also possibly might lead you into an unreadable memory page
Myeah but the BCL knows it's a string because of type safety
Pthe user has a duty of care to only use these when they're expecting a string
Kyeah
Kthis is unsafe stuff for sure
Mis a constptr meant to be a null terminated list
Pno, it's just a pointer to T
Pbut just like in C, a
byte* can be treated as a stringP(and friends)
Kits a
T* that doesnt need the unsafe keyword by the looks of itMbut with lifetime tracking
Pit's a ref struct, so the GC tracks it
Pit is marshalled to a string array to which we take a ref
Mmy only concern is that by doing unsafe operations when you compare a constptr to a string, that might be unexpected behavior
Min C you would use strcmp for this
Mand you know strcmp is unsafe
Mbut you do not know x == y is unsafe
Poh yeah ofc, to my knowledge
Span<T>.operator== does not compare the valuesPso we will follow the same precedent
P== is a reference equalsMso what is assert.that doing in there
Kthese tests arent == comparing the constptr
Kits doing a conversion back to string and then comparing that
Mahhhh
Pthat's just a string comparison
PthingBack is already a stringMis the conversion to string implicit or explicit
Kimplicit rn it looks like
Mshould be explicit, that's how C++ does it
Kid make it explicit personally just because of the unsafe-ty
Pimplicit for now, the idea with this is that it's magical with help of analyzers except where it is extremely dangerous
Pe.g.
string x = gl.GetString(GLEnum.Extensions) is acceptableMright that's what you want to do
Pbasically we're augmenting C# pretty heavily here to glue C# and C together, this will naturally have to include our own compile-time validation to hold our users hand
Pand when I mean augmenting C# I mean augmenting, in this MVP the
nullptr "keyword" existsMif i were making this i'd make glGetString return a special type like NullTerminatedConstPtr which is effectively the same thing but has that extra compiler metadata and that could have an implicit cast to string
Man analyzer would probably also work
Pgood food for thought though
Pi don't lie on either side of that, will have to think about it more
Pbut if one thing has come out of this conversation
Pis we NEED you for the next API review
Kaye
Mthat's fair
Pbut yeah I'm weirdly proud of
being a thing
being a thing
Plike it's completely horrible but it works
Mwhat does it do?
Pnullptr is from a global using static Silk.NET.Core.DSL; that returns a NullPtr which all pointer types can be implicitly casted fromPsounds inefficient but the JIT loves it and this is basically how C++'s nullptr works
Mit might not be a cleaner solution but i believe you can add an implicit cast from object and that will work with null
Mit just, yknow, has other problems