Maintenance PR + Hotfix

[CGMossa]

UPDATE: This PR got a little out of hand, so here's a description on what it does.

• get_non_api is only used when use-bindgen is enabled.
• Rename RCallbacks to reflect what they actually do right now, which is trimming the comments processed by clang, i.e. TrimCommentsCallbacks.
• BUG FIX: On Windows paths to dylib/dlls don't get processed properly. Added a .canonicalize and it seems to fix it.
• Updated bindgen to latest version 0.64.
• Added a versioning header to all bindings; e.g.

/* bindgen clang version: Ubuntu clang version 14.0.0-1ubuntu1 */
/* clang-rs version: Ubuntu clang version 14.0.0-1ubuntu1 */
/* r version: 4.1.3 */
/* automatically generated by rust-bindgen 0.64.0 */
[...]

• BUG FIX: Added a is_anonymous check on the entities processed. This is necessary as some unamed items
  gets a name in llvm16, and that isn't handled in any meaningful way by clang-rs unfortunately. This requires clang_3_7 atleast, which is also added.
• Updated bindings

I spent a lot of time inside the build.rs file and found a few things.
See the commit descriptions as well.

• Ensure generated bindings work with extendr/extendr. (i.e. no missing stuff please!)

[CGMossa]

I've ended up adding a hotfix for the issue that @Ilia-Kosenkov have been hunting down.
We've reported this to rust-lang/rust-bindgen#2488, and then I found out that it could potentially be the addition of unnamed structs in the named list of items in allowlist_pattern.

Needless to say, this should be removed in the future of course, once we get things patched upstream

[CGMossa]

Temporarily fixes #151

[yutannihilation]

Thanks for the great investigation! But, I see suspicious diffs on the bindings. I have no idea why the CI tests won't fail. Any ideas?

[CGMossa]

No I don't. But I don't like this either. I will export the `allowlist_pattern` lists and investigate if the allowlist suddenly changed and in turn no longer contains the "good" stuff. Unfortunately, this is somewhat cumbersome to do. But I'll get to it. Do you know if the `/bindings` command invokes CI? Why not?

…

On Tue, Apr 11, 2023 at 3:49 PM Hiroaki Yutani ***@***.***> wrote: Thanks for the great investigation! But, I see suspicious diffs on the bindings. I have no idea why the CI tests won't fail. Any ideas? — Reply to this email directly, view it on GitHub <#154 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIDVSA2NOG2ETCYSQPPVLTXAVONPANCNFSM6AAAAAAWZP4ZPM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[yutannihilation]

Do you know if the /bindings command invokes CI? Why not?

It doesn't invoke the CI because the commit message contains "[skip ci]".
The intent of this is that, as the bindings are the result of the CI runs, it should have failed if something is wrong.

And, sorry, I thought the CI checks with extendr so it would fail, but I was wrong here.

[CGMossa]

While this passes CI, running extendr on this PR doesn't pass. Example is R_BaseEnv is needed, but it got deleted (for some configurtion).

Thus I need to investigate this.. Sorry for the inconvenience.

[CGMossa]

Unfortunately, I do not have a path forward. Running this branch locally produces the bindings without the missing elements (atleast the one I sit on).
I can also not find where the logs for generating the bindings is at, so I'm unable to inspect what might have went on on the runner.

I definitely need assistance on this :(

[CGMossa]

No matter what I do, I'm unable to replicate the generated file on my machine with the one provided by the CI.

My version and the CI version agree on these

/* bindgen clang version: clang version 16.0.0 */
/* clang-rs version: clang version 16.0.0 */
/* r version: 4.2.3 */
/* automatically generated by rust-bindgen 0.64.0 */

and still there are some items that appear in my generated bindings locally, that are not present in the CI.

[yutannihilation]

Ah, so maybe adding " made the items on the head and tail of the allowlist unmatched?

b59b601

[CGMossa]

I lost a lot of my day yesterday to this little... issue.
I've updated the PR description and I kindly ask you to review and approve so this can be dealt with for now.

I think what we have right now is as much as an official fix will look like. Not just a hotfix.

[CGMossa]

/bindings

[Ilia-Kosenkov]

Amazing job, really well done! Thanks for also catching some annoying bugs, I'll definitely check it locally, especially whether the search path issue is resolved on Windows.

Feel free to resolve or reject the comments I left, they are mostly stylistic

[CGMossa]

@Ilia-Kosenkov can you confirm that I may merge this?

[yutannihilation]

Great work, thanks!