src: zero-initialize data that are copied into the snapshot

[joyeecheung]

This reverts one commit from #50983 because I found a better & simpler approach to prevent the padding from affecting snapshot reproducibility (in the second commit).

Revert "src: make sure that memcpy-ed structs in snapshot have no padding"

This reverts commit 4e58cde.

src: zero-initialize data that are copied into the snapshot

To prevent padding from making the snapshot unreproducible,
zero-initialize the data that are copied into the snapshot
so that the padding copied are all zeros. This is better
than enlarging the enums to align the fields since it doesn't
make the snapshot bigger than necessary, and it removes the
need of using static assertions to ensure alignment.

Refs: #50983

[nodejs-github-bot]

Review requested:

• @nodejs/startup

[joyeecheung]

cc original reviewers of #50983 @lemire @jasnell @legendecas

[joyeecheung]

For some reason, the value-initialize trick is not working in dynamically linked builds. So I removed the update and got the branch back to the original shape.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/59957/

[lemire]

For some reason, the value-initialize trick is not working in dynamically linked builds. So I removed the update and got the branch back to the original shape.

It should work as it is specified in the standard. The compiler basically calls memset for you, see https://godbolt.org/z/7hTW9norW

But the explicit memset is quite fine and, so we are clear, I did not request you change the code. :-/

I think that the code as it stands now is fine.

[joyeecheung]

The tests are still broken, so it may be something that landed in the main branch after cd8e61f that broke it (or could be the build infra?)

[joyeecheung]

Actually I am seeing the test failing on other unrelated PRs, so it could be either some infra update, or 296ce1b or 4c730ae

[richardlau]

Actually I am seeing the test failing on other unrelated PRs, so it could be either some infra update, or 296ce1b or 4c730ae

FWIW It started showing up in the reliability reports since yesterday: nodejs/reliability#905

[targos]

4c730ae may have exacerbated the issue, but not caused it. See #53583 (comment)