Hi Tudor,

some progress in Pharo 5:

> Am 28.07.2015 um 16:10 schrieb Tudor Girba-2 [via Smalltalk] <[hidden email]>:
>
> I also tried with Pharo4 and Pharo5, but it does not work because I get a MNU for GhostClassBuilder>>#client:.
>
> I fixed it in a Pharo5 image by using #installer: instead, but the computation of "JVM newWithSettings: jvmSettings" does not seem to finish and it blocks the VM.

Changing the call of #client: to #installer: in GhostClassInstaller>>#initialize: was correct.

> Any idea of what goes wrong?

Someone declared String>>subStrings:  as deprecated in Pharo 4 or 5. When I ran ���JVM newWithSettings: jvmSettings���, a notifier opened which warned about the deprecated method. I changed the method which used it in the debugger, then proceeded. The image blocked until I hit <cmd-.>. A debugger opened, it was in the middle of writing to the Pharo error log file. I proceeded, the image blocked again, I hit <cmd-.> again. A debugger opened which contained a very large stack of contexts with the method handleSignal: calling itself endlessly. There seems to be something wrong with handling/logging deprecation warnings in Pharo 5. Why do they have to be written to the error log in the first place? I don���t think that they belong there.

There are four methods sending subStrings: in the package JNIPort-Java-Base. I changed them to use substrings: instead.

There are two methods in JNIPort-StringEncoding which send "Smalltalk isLittleEndian���, String>>#asJavaLangStringEncodedByteArray and String>>#fromJavaLangStringEncodedByteArray:. This does not work anymore. I replaced this with "EndianDetector isLittleEndian���.

After these changes, starting a JVM and executing the example script which you posted just worked fine.

Unfortunately, loading the test package JNIPort-Tests-Java-Base fails with an exception:

UTF8InvalidText(Object)>>subclassResponsibility
UTF8InvalidText(Exception)>>defaultAction
UndefinedObject>>handleSignal:
Context>>handleSignal:

AFAIK, there was no problem when loading the package in Pharo 3. There seem to be two problems: The UTF8TextConverter seems to think that there is an invalid character in the source code file, and UTF8InvalidText has no #defaultAction method. Looks like a bug in Pharo 5 to me.

HTH
Joachim



View this message in context: Re: troubles with JNIPort on Pharo 5 / Java 8
Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.