Bug 16139 – REGRESSION: Number sign displayed inconsistently in URLs

• Status: NEW
• Component: Platform

[Alexey Proskuryakov]

"data:text/html,★☆" is displayed as "data:text/html,★&%239734;".

* STEPS TO REPRODUCE
1. Paste "data:text/html,★☆" into Safari address bar (no quotes).
2. Press Enter.

* RESULTS
The URL changes to "data:text/html,★&%239734;". It shouldn't change; or at least, the '#' character should be encoded in both cases.

* REGRESSION
This works fine in Safari 2 with r12006 WebKit, fails in Safari 2 with 10.4.11 WebKit. Also fails in Safari 3.

[David Kilzer (:ddkilzer)]

<rdar://problem/5619303>

[Darin Adler]

Is there an important symptom of this? This seems pretty minor to me.

[Alexey Proskuryakov]

None that I know of.

I looked into this briefly, and couldn't determine whether this was a WebKit of CF (or even Safari) bug. Or even if it's a bug or improved behavior :-/

[Darin Adler]

For what it's worth, the difference between the first number sign and subsequent ones is that the first is the one that separates the URL from the fragment identifier. The subsequent number signs are within the fragment identifier.

The issue is definitely in WebCore's KURL::parse function, not in Safari or CF.

The escaping is triggered by this entry in the characterClassTable:

    /* 35  # */ PathSegmentEndChar | BadChar

I suspect that the change in behavior for data URLs is due to http://trac.webkit.org/projects/webkit/changeset/24777>. That change caused us to treat "#" characters as the start of a fragment even in non-hierarchical URLs.

[Ahmad Saleem]

@Alexey - I am unable to reproduce this bug in Safari 17.5 Beta and copy pasting 'data:text/html,★&#9734' in URL bar and hitting Enter still keeps same URL.

I think this might have been resolved over the period of time and we can mark this as 'RESOLVED CONFIGURATION CHANGED'.

[Alexey Proskuryakov]

I'm not just seeing "data:" in Safari, despite the setting to show the full URL.