Minor simplification of LinkTest

[HannesWell]

And investigate why the test is failing.

Fixes #561

[github-actions]

Test Results

       42 files  ±0         42 suites  ±0   48m 25s ⏱️ - 7m 32s
  3 754 tests ±0    3 750 ✔️ +1    4 💤 ±0  0 ❌ ±0 
11 265 runs  ±0  11 235 ✔️ +1  30 💤 ±0  0 ❌ ±0 

Results for commit 6c8adc9. ± Comparison against base commit c5416ba.

♻️ This comment has been updated with latest results.

[HannesWell]

The offending link is probably on <help-URL>/help/topic/org.eclipse.platform.doc.isv/reference/extension-points/org_eclipse_help_index.html
My guess is that the failure of the LinkTest is caused by a space in some Extension-Point documentation in JDT.
Just created eclipse-jdt/eclipse.jdt.debug#278 to hopeful fix it.

The help-sites are

• https://help.eclipse.org/2023-06/index.jsp?topic=%2Forg.eclipse.jdt.doc.isv%2Freference%2Fextension-points%2Forg_eclipse_jdt_launching_vmInstalls.html&cp%3D3_1_1_20
• https://help.eclipse.org/2023-06/index.jsp?topic=%2Forg.eclipse.jdt.doc.isv%2Freference%2Fextension-points%2Forg_eclipse_jdt_launching_vmInstallTypes.html&cp%3D3_1_1_21

[HannesWell]

Even after eclipse-jdt/eclipse.jdt.debug#278 is merged and one I-build has passed the space in beginning is still found.
Stripping the string solves that problem and lets that test pass, but of course it would swallow similar problems in the future.

Does anybody know from which repo the IWorkbenchHelpSystem returned by PlatformUI.getWorkbench().getHelpSystem() is feed? Do I have to wait for another I-build, for the next milestone or even next release to get the updated version?

[Bananeweizen]

I suggest merging your change and not trying to find the original HTML document. AFAIK blanks around URIs are valid in HTML: https://www.w3.org/TR/2014/REC-html5-20141028/infrastructure.html#valid-url-potentially-surrounded-by-spaces

[Bananeweizen]

If I get it right, the source containing the blank has already been fixed in #561.

[merks]

If I get it right, the source containing the blank has already been fixed in #561.

Yes that's right. We did track down the offending files.

[Bananeweizen]

Sorry @HannesWell. I misread your code during first review. We still need to trim the regex result here to deal with URIs starting/ending with blanks.

Suggested change

	URI href = URI.create(matcher.group(1).replace(" ", "%20"));
	URI href = URI.create(matcher.group(1).strip().replace(" ", "%20"));

[HannesWell]

I suggest merging your change and not trying to find the original HTML document. AFAIK blanks around URIs are valid in HTML: https://www.w3.org/TR/2014/REC-html5-20141028/infrastructure.html#valid-url-potentially-surrounded-by-spaces

Thanks Michael for pointing this out, but unfortunately the help still uses html-4 as mentioned in #561 (comment). 😞

But I wonder why my changes in the Extension-Point schema are not picked up in an I-build when building the doc?
It looks like the doc is build in the I-builds:
https://github.com/eclipse-platform/eclipse.platform.releng.aggregator/blob/92d943bcd4eeba3187e64fc4f154ab029e234364/eclipse.platform.common/bundles/org.eclipse.jdt.doc.isv/buildDoc.xml#L33

Therefore I assumed the change fix is now included, but I still se it in my latest I-build SDK.

[Bananeweizen]

Thanks Michael for pointing this out, but unfortunately the help still uses html-4 as mentioned in #561 (comment). 😞

I'm absolutely not an expert on those standards, but looking at the differences between 4 and 5, the href attribute has no relevant changes: https://www.w3.org/TR/html5-diff/#changed-attributes. Therefore URIs starting/ending with blanks should be valid there, too.

[Bananeweizen]

Seems my knowledge of the standard is not sufficient enough (given the arguments in the other thread), so I'm removing my veto.

[HannesWell]

Thanks Michael for pointing this out, but unfortunately the help still uses html-4 as mentioned in #561 (comment). 😞

I'm absolutely not an expert on those standards, but looking at the differences between 4 and 5, the href attribute has no relevant changes: https://www.w3.org/TR/html5-diff/#changed-attributes. Therefore URIs starting/ending with blanks should be valid there, too.

Me neither. But in doubt, I would prefer to be stricter in this case, just to be sure.
And with eclipse-platform/eclipse.platform.releng.aggregator#1209 the fix in the jdt-doc should be picked up in tonights I-build and the failures should be gone.

[HannesWell]

Test is green now.