On Fri, 5 Jan 2024 13:52:23 GMT, Pavel Rappo <[email protected]> wrote:
> This PR improves diagnostic output and behaviour of the
> `TestMethodCommentsAlgorithm` test in environments that do not meet that
> test's expectations.
>
> The test assumes that a correct source file for `java.lang.Object` is
> available nearby. While the test verified that assumption, it didn't verify
> it deep enough. The issue described in the linked bug report seems to be that
> a directory that looks like a source directory does not contain that source
> file.
>
> The solution is to check the assumptions more thoroughly. Note that if the
> assumptions aren't met, the test will be skipped, but it will not fail. Some
> tools display a skipped test as **passed**, which could be misleading. If I
> were the original bug reporter, I'd investigate why the source file for
> `java.lang.Object` is missing.
>
> As for exception messages, I tried my best to make them helpful. That said,
> test exception messages are not user-level error messages. The stacktrace of
> an exception is supposed to be analysed in conjunction with the source that
> threw that exception.
test/langtools/jdk/javadoc/doclet/testMethodCommentAlgorithm/TestMethodCommentsAlgorithm.java
line 395:
> 393: String testSrc = System.getProperty("test.src");
> 394: if (testSrc == null) {
> 395: // shouldn't happen
Purely FYI, it could happen if the class is run directly, without `jtreg`.
That being said, very few people do that these days.
No change necessary,
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/17280#discussion_r1443390237
