It looks like the conversion from NSURL -> API::URL -> WTF::URL -> WTF::String -> JSValue is not going as expected. Since we don't examine or modify the URL, it suffices to convert it to NSURL -> WTF::String -> JSValue.
<rdar://problem/87357027>
Created attachment 448788 [details] Patch v1.0
(In reply to BJ Burg from comment #0) > It looks like the conversion from NSURL -> API::URL -> WTF::URL -> WTF::String -> JSValue is not going as expected. This seems like a bug, no? Or is this expected?
Created attachment 448802 [details] Patch v1.1
Comment on attachment 448802 [details] Patch v1.1 Looks good. Please note that using the URL class’s relative URL resolution doesn’t necessarily handle all the edge cases correctly and functions like Document::completeURL end up having to add special cases. So please think further on that point, for things like empty string.
(In reply to Darin Adler from comment #5) > Comment on attachment 448802 [details] > Patch v1.1 > > Looks good. > > Please note that using the URL class’s relative URL resolution doesn’t > necessarily handle all the edge cases correctly and functions like > Document::completeURL end up having to add special cases. So please think > further on that point, for things like empty string. Good point. In this case the code that provides the NSURL via ensures that the URL is non-nil and is not the empty string.
Comment on attachment 448802 [details] Patch v1.1 ios-wk2 and wincairo failures look to be infrastructure issues.
Committed r287895 (245933@main): <https://commits.webkit.org/245933@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 448802 [details].