mixed charset regression

[coolo]

I have a test suite that worked with bca84f9 (last commit last year) but fails with master

<"Invalid package name: 'Testing包صفقة'"> expected but was
<"Invalid package name: 'Testing包">.

The HTML queried is this:

Invalid package name: 'Testing包صفقة'

and first(:css, "div#flash-messages p").text is "Testing包"

Before I fail again while reducing the test case, I better report it as I see it this time :)

[coolo]

webkit_debug does not reveal anything unexpected:

Started "Node.text"
Finished "Node.text" with response "Success(Invalid package name: 'Testingå)"
Wrote response true "Invalid package name: 'Testingå"

[coolo]

the reason for this regression is rather suprising to me, but this "fixes" the problem:

--- a/src/WebPage.cpp
+++ b/src/WebPage.cpp
@@ -70,7 +70,7 @@ void WebPage::loadJavascript() {
}

void WebPage::setUserStylesheet() {

• QString data = QString("*, :first-line, :first-letter, :before, :after { font-family: 'Arial' ! important; }").toUtf8().toBase64();
• QString data = QString("*, :first-line, :before, :after { font-family: 'Arial' ! important; }").toUtf8().toBase64();
  QUrl url = QUrl(QString("data:text/css;charset=utf-8;base64,") + data);
  settings()->setUserStyleSheetUrl(url);
  }

[coolo]

reproducible in chromium - with that stylesheet .innerText is only the part without arabic. I changed my test case now not to include a line break :)

[panozzaj]

It looks like changes related to this might have been introduced in #442

[panozzaj]

I had the same issue with some tests of mine that worked fine locally, but when I pushed to tddium, I would get errors like: Expected to find "Apples" within text "pples ananas herries"

Where the text usually is "Apples Bananas Cherries". Basically it looked like the driver was disregarding the first and/or capitalized letters. I have a fork that has the change @coolo mentioned, as well as deleting the spec that it referenced. I'm not sure if I need to make a pull request. I couldn't get the specs running locally, but I think it should pass.

[mhoran]

I've seen a few reports of issues with #442, however aside from this mixed character set regression (which is odd, but likely an issue with the underlying implementation of innerText and something that can be worked around), I've been unable to reproduce.

As reported in #435, it seems that :first_letter is causing segfaults on OS X, however this does not happen for me. I have not been able to come up with a test case that causes a segfault.

Similarly, I have not been able to come up with a test case that causes the first letter to disappear. If you could provide such a failing test case, that'd be great. I think it'd be best to revert #442, however without test cases there's nothing to prevent a regression.

Also, reverting #442 will cause segfaults for OS X users with web fonts. The only workaround for this is to upgrade to Qt 5, as the upstream Qt bug report has not gotten any attention.

[mhoran]

I've released 0.14.1 to address this issue. We now only override web fonts for ::before and ::after pseudo elements and leave ::first-letter and ::first-line as is. This could cause segfaults for OS X users with web fonts applied to those pseudo elements, but that seems like less of an issue than the breakage reported here.

[mhoran]

@coolo, has 0.14.1 resolved this issue for you?

[coolo]

I changed my test case not to trigger the problem, so I don't know. I can check tomorrow (if I don't forget)

[mikegee]

@coolo 7 months have passed. Can you confirm that this issue is resolved?

[coolo]

verified: openSUSE/open-build-service@9e2f0f2