You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
@placeholder no longer a candidate for acc description #62
@stevefaulkner It's challenging to update this since it requires a patch to both the website and the test files which are in another repository. What to you recommend? Updating the results here and filing a new issue to the TPG repo?
As for this repo, the site fix seems to be: mark Safari as passing and mark Edge as failing. But I'd prefer to avoid that political pitfall unless @dstorey gives his blessing on the Edge edit. Otherwise, I'll just mark Safari as passing.
@jasonkiss@stevefaulkner this was correct but looks like the spec changed after we implemented :(
Is placeholder just removed from accDesc or also accName too? (presumably just the former)
If the former the fix is either to remove ph-002 and ph-003 tests or to leave them to explicitly test against the previously correct behavior and change the test desc to say placeholder shouldn't be the accDesc.
Update: Regarding Edge support, we're updating to remove placeholder from the name/description mapping to reflect the latest version of the spec.
@stevefaulkner how would you like the test to change? I can spare some cycles to update it if you like. My initial thought is to just remove the name/description requirement in the placeholder test, as it is consistent with the rest of the test where only features that contribute to name/desc algorithm include that test. The test already includes the requirement to map placeholder/aria-placeholder correctly.
Activity
cookiecrook commentedon Jan 23, 2018
@stevefaulkner It's challenging to update this since it requires a patch to both the website and the test files which are in another repository. What to you recommend? Updating the results here and filing a new issue to the TPG repo?
As for this repo, the site fix seems to be: mark Safari as passing and mark Edge as failing. But I'd prefer to avoid that political pitfall unless @dstorey gives his blessing on the Edge edit. Otherwise, I'll just mark Safari as passing.
cookiecrook commentedon Jan 23, 2018
#72 Updates Safari results
dstorey commentedon Feb 5, 2018
@jasonkiss @stevefaulkner this was correct but looks like the spec changed after we implemented :(
Is placeholder just removed from accDesc or also accName too? (presumably just the former)
If the former the fix is either to remove ph-002 and ph-003 tests or to leave them to explicitly test against the previously correct behavior and change the test desc to say placeholder shouldn't be the accDesc.
dstorey commentedon Feb 5, 2018
Regarding Edge implementation, if this is the new behaviour I'll work with the team to get Edge behaviour updated for an upcoming release.
dstorey commentedon Feb 16, 2018
Update: Regarding Edge support, we're updating to remove placeholder from the name/description mapping to reflect the latest version of the spec.
@stevefaulkner how would you like the test to change? I can spare some cycles to update it if you like. My initial thought is to just remove the name/description requirement in the placeholder test, as it is consistent with the rest of the test where only features that contribute to name/desc algorithm include that test. The test already includes the requirement to map placeholder/aria-placeholder correctly.
stevefaulkner commentedon Feb 23, 2018
@dstorey yeah if you can update would be great! the change you suggest sounds good.
dstorey commentedon Mar 10, 2018
Please see PR at ThePacielloGroup/AT-browser-tests#62