Combine fields sharing the same length-field into a single builder function

[gadunga]

So right now resolve_attachments takes in a slice of AttachmentReference, and then sets color_attachment_count to the len of the input slice. However according to the specification for VkSubpassDescription it states:

pResolveAttachments is NULL or a pointer to an array of colorAttachmentCount VkAttachmentReference structures defining the resolve attachments for this subpass and their layouts.

This seems to imply that the behavior of the builder function is incorrect in overwriting color_attachment_count. At the very least we should probably change it so that it doesn't do that. However I believe in the long run the correct behavior would be to verify that the len of the input slice matches color_attachment_count and produce a result of some kind.

[gadunga]

To be clear I'm willing to address this. Just let me know what course of action I should take. Thanks!

[MarijnS95]

Arrays are an often returning and discussed subject, because of their issues with optionality (or rather, lack of such annotations in the bindings), ambiguity for zero-length slices (such pointers should not be dereferenced) and the issue here: length fields shared between multiple pointers. For these cases I've been thinking about generating single functions that accept all involved slices at once, and asserts they're either identical in size or "null" (modeled as None or &[], haven't decided yet). The current separation is a bit too error-prone IMO.

To be clear I'm willing to address this. Just let me know what course of action I should take. Thanks!

Concrete answer: let's use this issue to discuss the appropriate way forward. I think single function for all slices could work, but others may have different solutions or find the drawbacks too severe.

[Ralith]

I've been thinking about generating single functions that accept all involved slices at once, and asserts they're either identical in size or "null"

I like this idea a lot, personally!

[MarijnS95]

@Ralith @MaikKlein Shall we rename this issue to something like "Combine fields sharing the same length-field into a single builder function" (or something more generic yet irrespective of SubpassDescriptionBuilder) and use it as a tracking issue to investigate such approaches?