Add tests for known violations of invariants

[jugglinmike]

JavaScriptCore and V8 have both been shown to violate an invariant of
the [[GetOwnProperty]] internal method in a number of cases. Add tests
asserting that the invariants are honored consistently.

Resolves gh-649. @claudepache I've listed your name in the tests' "author"
field. Please let me know if you're happy with that level of attribution!

[claudepache]

For the case of RegExp.$1, I leave to your judgment if including the test is relevant, knowing that

• according to my tests, of the four major JS engines used in web browsers, only WebKit fails it, falling short of your proposed criteria in Criteria for test relevancy #650;
• there is some chance that RegExp statics will be included in Annex B in the form of accessor properties (as part of https://github.com/claudepache/es-regexp-legacy-static-properties, stage 0 proposal to be championed by @erights in the next meeting), making that test obsolete.

The tests about Function#arguments and Function#caller seem ok to me, and definitely worth including.

Resolves gh-649. @claudepache I've listed your name in the tests' "author"
field. Please let me know if you're happy with that level of attribution!

No opposition to the attribution.

In case that copyright or authors' right is applicable to the snippet of code I've written in #649, I license it under CC0. Now, you might want to consult your lawyer to see whether it is correct to cite me as the sole author and at the same time attribute the copyright to the V8 project authors I’m not part of. (Not that I care, really.)

[leobalter]

In case that copyright or authors' right is applicable to the snippet of code I've written in #649, I license it under CC0. Now, you might want to consult your lawyer to see whether it is correct to cite me as the sole author and at the same time attribute the copyright to the V8 project authors I’m not part of. (Not that I care, really.)

I need to review this, but AFAIK, the content of test262 - as the ecma262 specs - becomes public domain or at least royalty free. So we might have a conflict here with the CC0 licensing.

[claudepache]

I need to review this, but AFAIK, the content of test262 - as the ecma262 specs - becomes public domain or at least royalty free. So we might have a conflict here with the CC0 licensing.

Note by CC0, I meant in practice as near as possible to public domain (but all juridictions do not allow to put into public domain).

[leobalter]

For conformance with the test262 project, the test must be offered in the same BSD license found here: https://github.com/tc39/test262/blob/master/LICENSE

We also need an agreement to the ECMA International RF Patent Policy and Ecma International Software Copyright Policy, see more at this link: https://github.com/tc39/ecma262/blob/master/CONTRIBUTING.md

Plus, we need that every contributor register to the TC39's CLA, http://www.ecma-international.org/memento/register_TC39_Royalty_Free_Task_Group.php

If you prefer, you can bring an own PR with this file content setting your own Copyrights (still BSD licensed).

Otherwise, we'll be stuck without being able to merge this, unless the same specific test is provided by someone that does not read this PR.

[rwaldron]

@claudepache ping

[leobalter]

@claudepache @bterlson if you're both attending this meeting, please take a look on this one. There'll be a cake.

[claudepache]

Sorry for the delay.

I'll definitely rewrite tests for function(){}.arguments and .caller with proper copyright and licence, as those properties are still buggy today in Chrome, Safari and Edge.

For RegExp.$1, I've precisely noted two distinct violations of invariants, each one in only one major browser:

• In Safari, the value of RegExp.$1 will incorrectly change under successful regexp matching, although it is reported as non-writable, non-configurable;
• In Edge, making RegExp.$1 non-configurable and non-writable will pretend to succeed, but the property is still reported as writable.

Should I write the tests for those two cases—knowing that they'll become obsolete if/when https://github.com/tc39/proposal-regexp-legacy-features is implemented (RegExp.$1 would become an accessor property)?

[erights]

Yes, please write those tests. That way we'll know more reliably when these problems do go away. Thanks.

…

On Tue, Jan 24, 2017 at 1:14 PM, Claude Pache ***@***.***> wrote: Sorry for the delay. I'll definitely rewrite tests for function(){}.arguments and .caller with proper copyright and licence, as those properties are still buggy today in Chrome, Safari and Edge. For RegExp.$1, I've precisely noted two distinct violations of invariants, each one in only one major browser: - In Safari, the value of RegExp.$1 will incorrectly change under successful regexp matching, although it is reported as non-writable, non-configurable; - In Edge, making RegExp.$1 non-configurable and non-writable will pretend to succeed, but the property is still reported as writable. Should I write the tests for those two cases—knowing that they'll become obsolete if/when https://github.com/tc39/proposal-regexp-legacy-features is implemented (RegExp.$1 would become an accessor property)? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#653 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAQtzOMnWSaOkWkbp1uh2AjAyU6p7lenks5rVmmzgaJpZM4Io22D> .

-- Cheers, --MarkM

[leobalter]

@jugglinmike with #841 now merged do you think there's anything else from here to ship? Please reopen if necessary.