feat: Implement Rust-side const ExternalOneByteStringResource subclass

[aapoalas]

MSVC and Itanium C++ ABIs agree that for simple inheritance the basic structure of a vtable contains metadata fields at a "negative offset" from the vtable pointer, and at zero or positive offsets come the virtual function pointers in the order of declaration. The only difference between the two is that MSVC only places the virtual deleting destructor in the vtable while Itanium ABI places both the deleting and the complete object destructors in it, leading to a vtable that is one pointer larger in Itanium / on Linux. Also MSVC only has a single metadata field instead of two for Itanium. Itanium inlines the base offset into the vtable while MSVC keeps it in what is essentially the entry point into the type info data.

Since the two are so similar, creating a custom vtable on Rust-side is pretty easy and can be done entirely at compile-time, meaning that instances of the class can also be created entirely at compile time. This leads to fully const external strings being possible.

The only thing that would be nice to add is some C++ side assertions for the vtable offsets but while member pointers to member functions are a thing, getting their offset is non-trivial. It's also a bit unclear to me if it's even possible to perform offset calculations with those.

[CLAassistant]

All committers have signed the CLA.

[mmastrac]

LGTM. @bartlomieju I think we should land and do a small experiment in Deno to confirm that it works at scale (ideally in a pretty heavily used codepath).

[bartlomieju]

Feel free to do it! I intend to cut a new release of rusty_v8 next week.

[avindra]

FWIW, the 32-bit build started failing here:

   ../../../src/binding.cc:68:1: error: static assertion failed due to requirement 'sizeof(v8::Isolate::DisallowJavascriptExecutionScope) == sizeof(unsigned int) * 2': DisallowJavascriptExecutionScope size mismatch
   static_assert(sizeof(v8::Isolate::DisallowJavascriptExecutionScope) ==
   ^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   ../../../src/binding.cc:68:69: note: expression evaluates to '12 == 8'
   static_assert(sizeof(v8::Isolate::DisallowJavascriptExecutionScope) ==
                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
   ../../../src/binding.cc:1033:5: error: static assertion failed due to requirement '__builtin_offsetof(ExternalConstOneByteStringResource, _length) == 16': ExternalConstOneByteStringResource length was not at offset 16
       static_assert(offsetof(ExternalConstOneByteStringResource, _length) == 16,
       ^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   ../../../src/binding.cc:1033:73: note: expression evaluates to '8 == 16'
       static_assert(offsetof(ExternalConstOneByteStringResource, _length) == 16,
                     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~
   ../../../src/binding.cc:1035:5: error: static assertion failed due to requirement 'sizeof(ExternalConstOneByteStringResource) == 24': ExternalConstOneByteStringResource size was not 24
       static_assert(sizeof(ExternalConstOneByteStringResource) == 24,
       ^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   ../../../src/binding.cc:1035:62: note: expression evaluates to '12 == 24'
       static_assert(sizeof(ExternalConstOneByteStringResource) == 24,
                     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~
   ../../../src/binding.cc:1037:5: error: static assertion failed due to requirement 'alignof(ExternalConstOneByteStringResource) == 8': ExternalConstOneByteStringResource align was not 8
       static_assert(alignof(ExternalConstOneByteStringResource) == 8,
       ^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   ../../../src/binding.cc:1037:63: note: expression evaluates to '4 == 8'
       static_assert(alignof(ExternalConstOneByteStringResource) == 8,
                     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~

[aapoalas]

@avindra Thanks for letting us know and sorry about the break! Can you check if this PR fixes the build on 32-bit architectures? #1289

[avindra]

Can you check if this PR fixes the build on 32-bit architectures? #1289

Sorry for the late reply. The 32 bit build passes now!