Update vnujar filtering to ignore no alt on role=img, again

[spectranaut]

There are more errors no on "multithumb-slider" related to the alt text, I update the regex to filter out all of the alt+img related errors. I wonder if we should open a bug on vnu-jar? Or are our examples actually not valid html?

Also this was fixed recently here: #1354

[spectranaut]

@nschonni can you review :) ?

[mcking65]

issue 429 is closed, and the related validator issue is closed too. Can we remove this?

[nschonni]

Looks like there are other hits now, so that issue is re-opened

[mcking65]

Do I have this right ... when updating vnu.jar from 20.3.16 to 20.6.30, the new version of vnu.jar finds errors related to the alt attribute on elements with role img that the prior version did not find?

Exactly what are the errors?

Do we have an open issue related to these errors to either fix them in the APG example or to track an open issue against the validator?

[nschonni]

@mcking65 there is an issue from @spectranaut here #1467 that has the new failure messages

[spectranaut]

I can't actually replicated this bug anymore after re-running npm install

[nschonni]

I was seeing it with the package bump here nschonni#8

[spectranaut]

Wow I'm really not seeing clearly today. I thought I was at the latest version of vnu-jar locally but I wasn't

[nschonni]

Wow I'm really not seeing clearly today. I thought I was at the latest version of vnu-jar locally but I wasn't

No worries, I know @sideshowbarker sometimes turns around patches pretty quick, so I checked to see that wasn't the case too :)

[sideshowbarker]

No worries, I know @sideshowbarker sometimes turns around patches pretty quick, so I checked to see that wasn't the case too :)

Is there some pending patch I haven’t merged? And that’s causing problems?

[mcking65]

In the multi-thumb slider, we have img elements with aria-label. If that is not valid, we need to change those to use alt. I didn't know that it is invalid to use aria-label instead of alt on img elements.

@sideshowbarker is that a recent change? In ARIA, aria-label is global. And, in ARIA, aria-label is supported on elements with the img role. So, whouldn't it be supported on elements with an implied img role?

[sideshowbarker]

@sideshowbarker is that a recent change?

I don’t know — I’d need to see the exact error or warning messages the HTML checker is emitting. I don’t see any actual error messages cited in this issue.

In ARIA, aria-label is global. And, in ARIA, aria-label is supported on elements with the img role. So, whouldn't it be supported on elements with an implied img role?

I don’t know. I’m happy to test with any actual markup for which the HTML checker doesn’t seem to be doing the right thing.

I do know that a while back (last year, I guess), I added a discretionary warning for aria-label on elements that the ARIA working group has been considering to disallow it for. Maybe that warning is what’s being run into in this case?

[mcking65]

@sideshowbarker

The errors are listed in #1467. We get the following errors on the multithumb slider example.

• An “img” element with no “alt” attribute must not have a “role” attribute.
• An “img” element with no “alt” attribute must not have any “aria-*” attributes other than “aria-hidden”.
• An “img” element must have an “alt” attribute, except under certain conditions. For details, consult guidance on providing text alternatives for images.

[spectranaut]

fixed those new comments, @nschonni ! :)

[nschonni]

@mcking65 this is the PR that will fix the vnu-jar version, and includes the config fixes

[nschonni]

@mcking65 this can land before my other PR to switch to Actions

[mcking65]

Unfortunately, as a screen reader user, It's really difficult to use the github actions output on the checks tab when trying to find what failed. In the navigation where it lists all the checks, it's really hard to find the one that failed. I think I figured out that the button for expanding has a badge with the number of actions with failures. It's extremely difficult to hear that badge; the label doesn't say it is the numbering of failing actions. Even harder is reading the log. It seems to be in a scrolling div of some kind and the scrolling doesn't work with a screen reader. Since the output is TAP, I used the search box to search for "not OK". That helped get the right content on the screen, but it was difficult to read.

All I figured out from the log is there is some kind of webdriver error.

[nschonni]

@mcking65 yeah, it looks like an unrelated regression test timeout issue. I think you might have an option to "Re-run check" for any failing tests, but I think this is OK to land either way.

There is a separate issue for seeing if we can add those inline failures like the JS/CSS/Spelling checks over here #1521

[nschonni]

Horray! All green again

[spectranaut]

@mcking65 it was an webdriver error... I think the answer really is to rerun the checks when that happens, but it's frustrating that it is so hard for you to get through to the content of the error from the github checks. I just wrote back to the internal Github Support tracker linking to your comment, for what it's worth!