fix: ensure v8 pointer compression + sandbox is enabled on 64bit native modules

[MarshallOfSound]

We screwed up in #34157 and created logic that said "turn off pointer compression, unless we are on a 32 bit host, in which case also turn off pointer compression" 😭

This also takes the opportunity to align other new flags in E20 (specifically sandbox related flags) with their state in our v8 build. Without this alignment everything just falls apart randomly.

Notes: Fixed spontaneous crashing in native modules that depended on nan

[deepak1556]

@MarshallOfSound the patch was removed in favor of the pointer compression being generated as part of config.gypi

electron/script/generate-config-gypi.py

Lines 17 to 19 in e83c3ec

	# Enabled in Chromium's V8.
	if target_cpu in ('arm64', 'x64'):
	args += ['--experimental-enable-pointer-compression']

[deepak1556]

I can confirm pointer compression is enabled with config.gypi headers from main https://output.circle-artifacts.com/output/job/c9303ce5-21b4-4005-a222-4c0c502b93fc/artifacts/0/node_headers.tar.gz

[VerteDinde]

@deepak1556 I don't think the way pointer compression is being generated now is working quite right, unfortunately. If you try to compile a native module that uses nan using any Electron 20 alpha or beta build, the module crashes.

You can see an example of that happening here (sorry, this example is Windows-only): https://gist.github.com/VerteDinde/90c7f4555c0f7ba5fd374fed3f20cbb0 Adding #define V8_COMPRESS_POINTERS above #include <node.h> within the native module itself fixes the crashing.

[MarshallOfSound]

@deepak1556 Ah I think I now understand the issue. We only upload a single node_headers bundle for each platform. But the value of enable_pointer_compression varies between architectures 🤔

This is why it works on CI for linux, because it uses the correct config.gypi 🤔 This change should probably still land because our current config.gypi has none of the sandbox flags and in fact has an incorrect one in one case in addition to fixing this pointer compression flag.

It should probably be considered a flaw in the generation of config.gypi if it generates arch-specific values without actually including a gypi arch check 🤔

[release-clerk]

Release Notes Persisted

Fixed spontaneous crashing in native modules that depended on nan

[trop]

I have automatically backported this PR to "20-x-y", please check out #34851

[MarshallOfSound]

We discussed this on Slack and concluded that if a slight adjustment (c556ce9) was made that this change was GTG