Allow components to be named the same as void tags #1390
Conversation
…rd HTML tags This includes a failing test case. If you parse to ast and immediately print back out, it turns: ```hbs <Link></Link> ``` into ```hbs <Link> ``` This behavior is special for `<Link>`. Other names don't do this. I suspect it's because, in strict HTML spec terms `<Link>` is the same as `<link>`, and `<link>` is supposed to not have a closing tag. But in Glimmer/Ember terms, `<Link>` is a component and dropping the closing tag introduces a syntax error. This was found in embroider-build/embroider#1176
…ts named the same as void elements
|
Per embroider-build/embroider#1176 (comment) it looks like there might be some other cases that might be worth testing / handling here? |
|
Makes sense to me! Updates pushed! |
…dability of the Printer
|
Hey, |
|
No, I'll chat with folks and see if we can get it into 4.8 as a bug fix (as well as the 4.9 beta releases). |
|
Is it possible to also hotfix this for older versions as well? Specifically 3.28? |
|
Almost certainly not. This is not a new bug, and it's definitely not a security bug (3.28 is in the security-only phase of its LTS cycle). I understand that it's annoying, and we will definitely get it onto 4.8 and possibly onto 4.4, but we will need to discuss even the 4.4 case. It would be one thing if this were a regression, but it is not, as far as I can tell, and we don't make a habit out of doing significant backports for general bug fixes, else we would end up having to do it for basically every bug fix. Don't get me wrong: I am sympathetic and it is annoying! But there are also costs to a backport: if we mess something up, we then have to revert or try to fix that, which causes yet more churn and difficulty for our users! And: there are only so many hours in the day. |
|
Yes makes sense. @bertdeblock do you think we can add a disclaimer for the lint rule or something? It's somewhat unusable right now no? |
|
@Alonski To fix the lint rule, I think we just need a new release here and then update the |
|
Oh that makes sense. I was thinking we had to update Ember source as well but what you said is that we don't. Thanks! |
|
Would it be possible for someone to already release this? Getting it in the appropriate Ember versions can still be done / discussed afterwards I believe? Having this available in a new release would already allow us to fix the issue over in |
|
I pinged a couple folks with permissions in Discord but I believe one is on vacation and the other had a sick kid this week. |
|
Comment via @wycats: there may still be a remaining bug here because if the tokenizer is not scope-aware, not-capitalized cases could still get wrong handling: |
Closes: #1388
Much thanks to @ef4 and @Windvis
tl;dr:
<Link></Link>should be an allowed component invocation.Why? ლ(ಥ Д ಥ )ლ
Background: embroider-build/embroider#1176
(though, @runspired isn't the first to run in to this, I've noticed this issue myself for the longest time, and just had a hard time understanding what was going on, because I was new into my "how do I even debug the things"-with-embroider-journey 😅
This is caused by this line.
Though, a bit of concern
I think long-term this is something that will fizzle out as a class of problem, because of strict mode (but who knows how long it'll be until everything is strict mode).
The drawbacks ┌༼▀̿ Ĺ̯▀̿༽┐
Anyone using
Link,LINK,LiNk,lINK,liNK,linKas their style/asset loader in the their glimmer-parsed template will now not be a void element.how many people are doing this tho? 🙃
The importance (ﾉ◕ヮ◕)ﾉ*:･ﾟ✧
Not being able to have a
<Link>component with embroider-using apps is a blocker for many many many many apps / addons I maintain from being able to adopt embroider.