On Wed, 20 Oct 2021 10:15:39 GMT, Hannes Wallnöfer <hann...@openjdk.org> wrote:
>> test/langtools/jdk/javadoc/doclet/testSnippetTag/TestSnippetTag.java line >> 214: >> >>> 212: <span >>> class="element-name">case%s</span>()</div> >>> 213: <div class="block"> >>> 214: <div class="snippet-container"><a href="#" >>> class="snippet-copy" onclick="cop\ >> >> I'm sorry for hard-coding the snippet HTML output; I should've encapsulated >> it instead. Now every change to the HTML structure ripples through the >> tests. I'll work on improving that, I promise. >> >> Separately. Aren't these "Copy" and "Copied!" strings localized? What >> happens if this test is run on some other locale? > > Yes, they are localized. Good question, I do believe tests are run with fixed > en/US locale, but I'm not absolutely sure. I've traced the execution of a local jtreg run with a debugger and examined the respective `jtr` file. Here is what I saw. The locale that the javadoc under test uses is that of returned by java.util.Locale.getDefault(). This happens because neither of the three pieces -- jtreg, JavadocTester, or TestSnippetTag -- passes `-locale` to javadoc. Now I begin to wonder: how come we don't see failures on CI builds? Are we lucky and all the machines on which javadoc tests run use variants of English locale? Do we not have other tests where we hard-code English string variants? Is there something that I don't see? ------------- PR: https://git.openjdk.java.net/jdk/pull/6011
