Add Opera for Android

[ExE-Boss]

Part of #591

Note: This might not actually be fully accurate, as I don’t use Opera.

review?(@Elchi3, @jpmedley)

[Elchi3]

So, this is tricky, but I found this: https://ftp.opera.com/pub/opera/android/

I can confirm these versions:

• 10.1, 2010-11-09, https://dev.opera.com/blog/opera-mobile-10-1-beta-for-android-is-here
• 11, 2011-03-22, https://dev.opera.com/blog/opera-mobile-11-for-maemo-meego-windows
• 11.1, 2011-06-30, https://dev.opera.com/blog/opera-mobile-11-1-new-features-and-additions
• 11.5, 2011-11-02 (?), src
• 12, 2012-02-25, https://dev.opera.com/blog/opera-mobile-12-and-introducing-opera-mini-next
• 12.1, 2012-10-09, https://dev.opera.com/blog/opera-mobile-12-1-with-spdy-web-sockets-flexbox-and-more
• 14, 2013-05-21, src, based on Chromium v26
• 15 2013-07-02, src1, src2, ???
• 16, 2013-09-18, src, based on Chromium v29
• 18, 2013-11-20, src, based on Chromium v31
• 19, 2014-01-22, src, based on Chromium v32
• 20, 2014-03-06, src, based on Chromium v33
• 21, 2014-04-22, src, based on Chromium v34
• 22, 2014-06-17, src, based on Chromium v35
• 24, 2014-09-10, src, based on Chromium v37
• 25, 2014-10-16, src, based on Chromium v38
• 26, 2014-12-15, src, based on Chromium v39
• 27, 2015-01-29, src, based on Chromium v40
• 28, 2015-03-10, src, based on Chromium v41
• 29, 2015-04-28, src, based on Chromium v42
• 30, 2015-06-10, src, based on Chromium v43
• 32, 2015-09-23, src, based on Chromium v45
• 33, 2015-11-02, src, based on Chromium v46
• 34, 2015-12-15, src, based on Chromium v47
• 35, 2016-02-03, src, based on Chromium v48
• 36, 2016-03-31, src, based on Chromium v49
• 37, 2016-06-15, src, based on Chromium v50
• 41, 2016-12-22, src, based on Chromium v54
• 42, 2017-02-20, src, based on Chromium v55
• 43, 2017-09-27, src, based on Chromium v59
• 44, 2017-12-11, src, based on Chromium v60
• 45, 2018-02-14, src, based on Chromium v61
• 46, 2018-05-14, src, based on Chromium v63
• 47, 2018-07-23, src, based on Chromium v66
• 48, 2018-11-08, src, based on Chromium v69
• 49, 2018-12-07, src, based on Chromium v70

So far I can't confirm:

• 17
• 23
• 31
• 38 - 40
• 49-60

[Elchi3]

Oh and I think in this case it doesn't make sense to link to Opera Desktop release notes as it is not the same.

[jpmedley]

It looks like I missed responding to a request for review. I do not have access to any information beyond the sources the Forlian has already consulted.

[Elchi3]

@ExE-Boss Would you like to update the Opera for Android data based on #1712 (comment)? I'm curious to see how much data will fail and how we could correct it.
This is the last browser without data/validation, and it is the next infra item I'd like to finish.

[ExE-Boss]

I plan to look into it during the weekend before v0.0.52 eventually.

[queengooborg]

Moving the discussion to this PR. What was discussed in #2909 is as follows:

Comment 1

Looks like we have a little bit of an inconsistency on our hands in terms of the version numbers. Opera Android has different version numbers than Opera Desktop, that’s already been established in the discussion on #1712 up above.

https://www.apk4fun.com/history/998/
https://en.wikipedia.org/wiki/Opera_Mobile

However, ChromeStatus.com seems to be referring to Opera Android version numbers as if they were the same as the desktop version numbers, as well as referring to versions way beyond the current released version.

https://www.chromestatus.com/feature/5718547946799104

So...it must be referring to the underlying engine in Opera Android, which follows the desktop version’s numbers?

Comment 2

I think the version numbers that I [@Elchi3] mention here #1712 (comment) are correct. Whether or not that's actually doable and makes sense for our data, I don't know yet. Strictly speaking chromestatus seems to do the wrong thing, but maybe they refer to some engine versions. I haven't had the time to dig deeper into this yet.

Comment 3

As far as version numbers, a similar case is evident with Microsoft Edge, as described in #2620, where the engine and the application versions are different. As also described, the engine version is preferred, rather than the application version. Personally, I feel that maybe we should keep with that regarding Opera Android (especially since I expect people would believe chromestatus.com's accuracy, and it would also be a little less work :P). But it's all up to you guys in the end, what you think would be best!

[jpmedley]

"Looks like we have a little bit of an inconsistency on our hands in terms of the version numbers. Opera Android has different version numbers than Opera Desktop, that’s already been established in the discussion on #1712 up above."

According to one of Chrome's Developer Advocates who worked for Opera until a few years ago:

The wikipedia article is incorrect. Opera for Android followed the OPERA_VERSION = CHROMIUM_VERSION - 13 formula, just like Opera for Desktop. However, it looks like this recently changed. The current Opera for Android version is v47~, but its Chromium version is v66~.

Opera did not tell us.

[Elchi3]

My wild guess is that "OPERA_VERSION = CHROMIUM_VERSION - 13" worked until Opera Android 37. I think it would be great to find out if there is a new formula for the later versions.

Edit: See above #1712 (comment) for what we currently know.

[mathiasbynens]

Opera employees should know.

[Elchi3]

@chrisdavidmills Do you know Opera employees that could help us here?

[chrisdavidmills]

@Elchi3 I don't know anyone from Opera any more; everyone I know left.

@mathiasbynens was there much more recently than me; so was Andreas Bovens.

[poshaughnessy]

In case a few more are helpful in the meantime...

• v45 = Chromium v61 (found from Play Store description from Internet Archive)
• v46 = Chromium v63 (found from this Opera Forums announcement, NB. possibly you could try contacting Miyuki there via the forum?)
• v47 = Chromium v66 (found from opera://version/ using current Stable)
• v48 = Chromium v69 (found from opera://version/ using current Beta)

[Elchi3]

Thanks @poshaughnessy! 👍 Especially the forum link was super helpful! I've updated my comment above (#1712 (comment)) with what we know now.

[Elchi3]

I've only reviewed files in api/ that start with "A" and I found things that look wrong or maybe I don't understand the changes. Please fix or explain.

[ExE-Boss]

review?(@Elchi3): I’ve now addressed your review.

Let’s get this merged before more stuff breaks.

[foolip]

There are a lot of conflicts already and this will also conflict with #3658.

If there's some vetting of the API/CSS/etc. data needed, can the browsers/opera.json and browsers/opera_android.json changes be merged first? Per #591 opera_android is the only browser missing in browsers/, and I noticed this today when trying to update data for it in #3731.

[ExE-Boss]

@foolip You need to run a script that updates the data according to #1712 (comment) (without it, the build will fail, as right now, the data assumes that Opera Android version == Opera Desktop version, but the Opera Android version is at least 9 versions behind).

[foolip]

@ExE-Boss I'm not setting any version numbers though, just changing null to true.

[ExE-Boss]

My script also affects the case when opera_android is null and opera is non‑null.

[foolip]

So, do I need to do anything if I'm just changing null to true based on testing in the real browser, or is it just that there's still more work that could be done?

[ExE-Boss]

It means that there will be merge conflicts (I’d recommend merging this first, as this addresses some version_added: nulls).

[queengooborg]

We'll definitely be facing merge conflicts, so the sooner we can get this merged, the better!

[Elchi3]

So, I've looked into this again and the mapping seems to be applied correctly now. Thanks for your work!

The only thing that this doesn't seem to take into account is that when mobile chromiums differ from desktop chromiums. This is quite rare, but for example, look at https://developer.mozilla.org/en-US/docs/Web/CSS/@viewport/zoom. Here, Chrome for Android is much later than Chrome desktop (I don't know if the data is correct btw) but your mapping here takes the Chrome desktop number. I would assume that Opera Android is closer to Chrome for Android and thus the Opera Android number could have been based on the Chrome Android version.

I believe that majority of the changes in this PR is correct and I think that the above scenario is relatively minor. So, I think it would make sense to go ahead and merge this if I don't hear other thoughts on this matter.

It needs another rebase though. Sorry.

[jpmedley]

Rule of thumb: if device hardware is involved, don't be surprised if desktop and mobile have different version numbers. In fact you can almost expect it. (The hardware interaction here comes from the fact that zooming on mobile has to be done by pinching the screen.)

The difference is usually only a few versions. There seems to be a difference of 32 versions for this feature. If the difference is excessively large, ping me and I'll look into it.

[Elchi3]

Thanks for the rebase!

I've double checked again https://github.com/mdn/browser-compat-data/blob/aa314c9bce0d3dea20a08011a0f6feeb833fc52a/browsers/opera_android.json against #1712 (comment)

In the comment I said that there weren't these releases:
17
23
31
38 - 40

But you've added them to opera_android.json. Why? Did you find evidence that these releases existed?

[ExE-Boss]

I think I forgot to remove them.

[Elchi3]

We need to remove the non-existing releases here.
Also, I think, linking to the desktop release notes is more confusing than helpful (this can be a follow-up, though).

[ExE-Boss]

review?(@Elchi3): I’ve removed the non‑existent Opera Android versions, now would probably be a good time to merge this before more merge conflicts pile up.

[Elchi3]

Okay, here we go, I'm merging this. 🎉
Thanks for your continued work on this for over a year, @ExE-Boss! 🥇 Very much appreciated 👍
This wasn't an easy one.