[foreign] RFR 8222919: jextract should compile generated java sources rather than use ASM to generate class files

[Jorn Vernee]

Looks good!

Now that the transition to source code generation is complete, and the 
ASM factories are removed, we can start looking at the plugin/extension 
story some more I think?

Cheers,
Jorn

Sundararajan Athijegannathan schreef op 2019-04-24 16:16:
> Hi,
> 
> Fixed. Updated webrev: 
> https://cr.openjdk.java.net/~sundar/8222919/webrev.01/
> 
> Issues:
> 
> 1) Wrong source file name for static forwarder classes -harmless (as
> the name is used only in SourceFile attribute) and hence no test
> failure because of that. But it has been fixed to use correct source
> file name (with package name rather than just simple name).
> 
> 2) '\\' in names of Zip/Jarfile:
> 
> Previously Map<String, byte[]> used "/" separated names for class
> names (com/acme/Foo for com.acme.Foo). This masked the bug in
> JarWriter (which replaced '.' with File.separatorChar - but "." didn't
> occur in the name at all)! With InMemoryJavaCompiler, names are normal
> fully qualified classnames (com.acme.Foo) - which JarWriter
> transformed with "\\" (File.separatorChar) on Windows! Resulting in
> zip/jar file containing '\\' chars! This is a bug per Zip standard.
> 
> https://pkware.cachefly.net/webdocs/casestudies/APPNOTE.TXT
> 
> "4.4.17.1 The name of the file, with optional relative path. The path
> stored MUST NOT contain a drive or device letter, or a leading slash.
> All slashes MUST be forward slashes '/' as opposed to backwards
> slashes '\' for compatibility with Amiga and UNIX file systems etc. If
> input came from standard input, there is no file name field. "
> 
> :)
> 
> PS. Jorn and I discussed this via private emails as well - found the
> same issue independently! Submitting internal mach5 job concurrently.
> 
> Thanks,
> -Sundar
> 
> On 24/04/19, 6:48 PM, Sundararajan Athijegannathan wrote:
>> Hi Jorn,
>> 
>> Yes, I see 3 test failures on Windows (only) with internal mach5 
>> build. I'll check this out. Thanks.
>> 
>> -Sundar
>> 
>> On 24/04/19, 4:35 PM, Jorn Vernee wrote:
>>> Hi Sundar,
>>> 
>>> In InMemoryJavaCompiler.FileManager::getJavaFileForOutput; Should 
>>> this use computeIfAbsent instead of put? You're probably more aware 
>>> of the backing implementation. Is there any chance a file with the 
>>> same name is requested twice, and then the previously created 
>>> ClassFile object being overwritten?
>>> 
>>> Also, there are some tests failing. This seems to be due to Unix vs. 
>>> Windows path separators, for instance in the Runner test:
>>> 
>>>   test Runner.testJarManifest(): failure
>>>   java.lang.AssertionError: Sets differ: expected [com.acme.pad_h, 
>>> com.acme.pad_h$anon$pad_h$1195, com.acme.pad_h$PaddyStruct] but got 
>>> [com\\acme\\pad_h$anon$pad_h$1195, com\\acme\\pad_h, 
>>> com\\acme\\pad_h$PaddyStruct]
>>>     at org.testng.Assert.fail(Assert.java:94)
>>> 
>>> I'm looking into this right now, but maybe you know where the problem 
>>> might be?
>>> 
>>> Cheers,
>>> Jorn
>>> 
>>> Sundararajan Athijegannathan schreef op 2019-04-24 11:34:
>>>> Please review.
>>>> 
>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8222919
>>>> Webrev: https://cr.openjdk.java.net/~sundar/8222919/webrev.00/
>>>> 
>>>> Thanks
>>>> -Sundar