Have test links link more directly to github test262 pages.

The old heuristic for dereferencing test: identifiers has not worked
for a long time. Change these from test ids to more direct links into
the github test262 repository. For brevity, still recognize when a
link starts with "test/", in which case prepend the github link
prefix.

Two tests use the full url because the tests have been removed from
github, and so we link to an historical version.

[kpreid2, 2015-02-15 22:15:54]

LGTM

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
File src/com/google/caja/ses/useHTMLLogger.js (right):

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
src/com/google/caja/ses/useHTMLLogger.js:203: return test;
I don't like that this as it is behaves oppositely to normal relative URI
resolution.

Given the circumstances, I suggest a condition at this point checking for
/^[-+.\w]+:/ (that is, has a URI scheme) and indicate failure (i.e. return
something like "data:,Failed%20to%20resolve%20" + test) otherwise. That'll at
least make it never return nonsense.

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
src/com/google/caja/ses/useHTMLLogger.js:258: link.href =
'http://es5.github.io/#x' + encodeURIComponent(section);
We can https: here

[MarkM, 2015-02-16 00:10:38]

The old heuristic for dereferencing test: identifiers has not worked
for a long time. Change these from test ids to more direct links into
the github test262 repository. For brevity, still recognize when a
link starts with "test/", in which case prepend the github link
prefix.

Two tests use the full url because the tests have been removed from
github, and so we link to an historical version.

[MarkM, 2015-02-16 00:11:09]

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
File src/com/google/caja/ses/useHTMLLogger.js (right):

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
src/com/google/caja/ses/useHTMLLogger.js:203: return test;
On 2015/02/15 22:15:54, kpreid2 wrote:
> I don't like that this as it is behaves oppositely to normal relative URI
> resolution.

What would you suggest instead?


> Given the circumstances, I suggest a condition at this point checking for
> /^[-+.\w]+:/ (that is, has a URI scheme) and indicate failure (i.e. return
> something like "data:,Failed%20to%20resolve%20" + test) otherwise. That'll at
> least make it never return nonsense.

Done.

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
src/com/google/caja/ses/useHTMLLogger.js:258: link.href =
'http://es5.github.io/#x' + encodeURIComponent(section);
On 2015/02/15 22:15:54, kpreid2 wrote:
> We can https: here

Done.

[kpreid2, 2015-02-16 00:14:31]

LGTM

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
File src/com/google/caja/ses/useHTMLLogger.js (right):

https://codereview.appspot.com/199650043/diff/1/src/com/google/caja/ses/useHT...
src/com/google/caja/ses/useHTMLLogger.js:203: return test;
On 2015/02/16 00:11:09, MarkM wrote:
> On 2015/02/15 22:15:54, kpreid2 wrote:
> > I don't like that this as it is behaves oppositely to normal relative URI
> > resolution.
> 
> What would you suggest instead?

My next paragraph, which you have already implemented, was my suggestion.