Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 12/11/2015 14:41, Wang Weijun wrote:

make/Images.gmk uses both JDK_COMPACTn and JRE_COMPACTn. Should it be only 
JRE_COMPACTn?


Yes, Images.gmk should be more consistent in the naming of its variables.

-Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 11/11/2015 11:07, Wang Weijun wrote:

:
I'm prototyping in jake and the change is

src/java.base/share/classes/module-info.java:

  exports sun.security.util to
+jdk.security.cert,
  exports sun.security.x509 to
+jdk.security.cert,

src/jdk.security.cert/share/classes/module-info.java:

+module jdk.security.cert {
+exports jdk.security.cert;
+}


Okay but is the jdk.security.cert API really tied to modules? If not 
then I assume this can be pushed via jdk9/dev and we can create the 
module declaration as part of sync'ing up the jake forest.


-Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Wang Weijun]

> On Nov 11, 2015, at 3:24 PM, Alan Bateman  wrote:
> 
> 
> On 11/11/2015 03:20, Wang Weijun wrote:
>> Ping again.
>> 
>> Will the future JRE only include java.se? In other words, what will the 
>> java.com download for non-developers contain?
>> 
> This isn't really a quick for here. However, in your build environment then 
> use jre/bin/java -listmods to see the full list of modules that are currently 
> linked in. In the makes files then /make/Images.gmk is the make file to 
> check out, specifically the MAIN_MODULES and PROVIDER_MODULES as these are 
> the set of root modules that will be resolved when creating the "jre" image.

Great, I see MAIN_MODULES contains quite some jdk.* modules.

Here comes the question: How do I request a new module to be included in it? 
The CCC for the new module/API?

Thanks
Max

> 
> -Alan.

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 11/11/2015 09:54, Wang Weijun wrote:

:

Here comes the question: How do I request a new module to be included in it? 
The CCC for the new module/API?

Can you cc jigsaw-dev on the changes to modules.xml? modules.xml is the 
temporary document that we have in JDK 9 to define the modular structure 
until the module system goes into the main line.


-Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 11/11/2015 03:20, Wang Weijun wrote:

Ping again.

Will the future JRE only include java.se? In other words, what will the 
java.com download for non-developers contain?

This isn't really a quick for here. However, in your build environment 
then use jre/bin/java -listmods to see the full list of modules that are 
currently linked in. In the makes files then /make/Images.gmk is 
the make file to check out, specifically the MAIN_MODULES and 
PROVIDER_MODULES as these are the set of root modules that will be 
resolved when creating the "jre" image.


-Alan.

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Wang Weijun]

> On Nov 11, 2015, at 5:03 PM, Alan Bateman  wrote:
> 
> On 11/11/2015 09:54, Wang Weijun wrote:
>> :
>> 
>> Here comes the question: How do I request a new module to be included in it? 
>> The CCC for the new module/API?
>> 
> Can you cc jigsaw-dev on the changes to modules.xml? modules.xml is the 
> temporary document that we have in JDK 9 to define the modular structure 
> until the module system goes into the main line.

I'm prototyping in jake and the change is

src/java.base/share/classes/module-info.java:

 exports sun.security.util to
+jdk.security.cert,
 exports sun.security.x509 to
+jdk.security.cert,

src/jdk.security.cert/share/classes/module-info.java:

+module jdk.security.cert {
+exports jdk.security.cert;
+}

--Max

> 
> -Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Wang Weijun]

Ping again.

Will the future JRE only include java.se? In other words, what will the 
java.com download for non-developers contain?

Thanks
Max

> On Nov 5, 2015, at 10:07 AM, Wang Weijun  wrote:
> 
> I was talking with Vincent off the list, but it seems better to be back.
> 
> Here is Vincent's mail on how he thinks of the new API we are about to 
> provide in JDK-8058778.
> 
>> The new API is interesting but if it's not available in Java SE I'm afraid 
>> it does not fit our use case. We cannot make JOSM depend on a JDK at runtime 
>> just for this feature :(
> 
> There might be some confusion on what "JDK" means. It used to be the 
> developer's toolkit versus JRE the runtime, but as for module names, a jdk.* 
> module contains APIs defined by OpenJDK instead of Java SE. In fact, most of 
> them are not especially designed for developers.
> 
> That said, I am not exactly sure what kind of images we will release after 
> jigsaw. If there is still a JRE, I think it will include not only the java.* 
> modules but also some jdk.* ones. I hope jdk.security.cert is one of them.
> 
> Personally I am not afraid of using jdk.* modules, since almost every JDK 
> distribution is based on OpenJDK. At least it is now a supported API of 
> OpenJDK instead of old sun.* classes that were never claimed to be supported 
> by anyone.
> 
> In fact, keytool was not defined in Java SE either. It was in JRE though.
> 
> Thanks
> Max

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[dalibor topic]

On 30.10.2015 16:29, Vincent Privat wrote:

Hi,
- Is it expected to allow external people to have the possibility to
subscribe to JDK issues?


See http://robilad.livejournal.com/139637.html

cheers,
dalibor topic

--
 Dalibor Topic | Principal Product Manager
Phone: +494089091214  | Mobile: +491737185961


ORACLE Deutschland B.V. & Co. KG | Kühnehöfe 5 | 22761 Hamburg

ORACLE Deutschland B.V. & Co. KG
Hauptverwaltung: Riesstr. 25, D-80992 München
Registergericht: Amtsgericht München, HRA 95603

Komplementärin: ORACLE Deutschland Verwaltung B.V.
Hertogswetering 163/167, 3543 AS Utrecht, Niederlande
Handelsregister der Handelskammer Midden-Niederlande, Nr. 30143697
Geschäftsführer: Alexander van der Ven, Astrid Kepper, Val Maher

 Oracle is committed to developing
practices and products that help protect the environment

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 30/10/2015 15:29, Vincent Privat wrote:

:
- Some issues didn't make it to the public JIRA and remained in the 
private bug database. Can we please have more information on them (why 
have they apparently been rejected)? The incident numbers 
are JI-9009025, JI-9010791, JI-9009449, JI-9008003.


Incidents submitted via bugs.sun.com have historically been hit or miss. 
I don't know why these four were not moved to the JDK bug database, but 
here they are now:


JI-9009025 => https://bugs.openjdk.java.net/browse/JDK-8141093

JI-9010791 => https://bugs.openjdk.java.net/browse/JDK-8141094

JI-9009449 => https://bugs.openjdk.java.net/browse/JDK-8141095

JI-9008003 => https://bugs.openjdk.java.net/browse/JDK-8141096

The OpenJDK mailing lists aren't a support channel but reproducible test 
cases are a wonderful way to contribute. So if you have test cases or 
patches that can be added to these bugs for the issues that are in 
OpenJDK then it's best to follow-up on the appropriate mailing list 
(likely awt-dev or swing-dev for some of these).




:

Finally:
- We had a terrible experience when trying to report a bug against 
JAXP. We detected a severe data corruption problem in StaX when 
dealing with Unicode SMP characters, so we reported it, including a 
sample Java program 100% reproducible, in January 2013 (JAXP-76 on 
java.net  JIRA). As no activity was visible on this 
JIRA instance, we tried to use the standard Java bug report, three 
times, without success, with incident numbers 2431783 
(2013-01-23), 2627098 (2013-10-28) and 9048481 (2014-11-28), without 
any answer. On 2014, November 29th we discovered by chance that the 
bug had finally been detected and fixed internally, as JDK-8058175 
(created and resolved in September 2014). We reported back to the 
public JAXP JIRA instance, again without any answer. 6 months later we 
finally got the ironic and laconic answer "Please report issues to the 
OpenJDK Bug System", which was exactly was we were trying to do for 2 
years! Can you please tell us why our bug reports were all silently 
ignored while the bug was real, and if is it still worth reporting 
bugs against JAXP? Thankfully we had far better experiences with other 
components of the JDK.
JAXP development moved from the jaxp.java.net project to OpenJDK in 
early 2012. The project main page should be clearer on this point. I 
wasn't aware there was a JIRA instance but it probably should be shut 
down to avoid confusion.


In any case, here is the mapping:

2431783 => https://bugs.openjdk.java.net/browse/JDK-8132652

2627098 => this one is on the old Sun web incident reporting system, I 
don't know how to move it to JIRA but it seems to be a dup of 
JDK-8141097 or JDK-8058175.


JI-9048481 => https://bugs.openjdk.java.net/browse/JDK-8141097

As you note, JDK-8058175 has been fixed in JDK 9 builds for some time 
(since jdk9-b31). If you can verify the fix with the JDK 9 builds then 
you could lobby on jdk8u-dev to have it back-ported to 8u.


-Alan.

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 31/10/2015 13:40, Vincent Privat wrote:

Hi Alan,
Thanks a lot for your answers and guidance. We will follow-up bugs 
with appropriate mailing lists and attach test cases every time possible.

Best regards,
Vincent

Thanks. So I think from the list of 29 issues then there are only 2 that 
are specific to the jigsaw EA builds. One of these is caused by the 
horrible hack to workaround JDK-6292739, I see you've started a thread 
on awt-dev about that issue. The other is the deadlock issue which is 
fixed for the next Jigsaw EA build.


-Alan.

JOSM feedback on Java 7,8,9, including Jigsaw EA

[Vincent Privat]

Hi,
Following the recent inclusion of JOSM to the Quality Outreach list by the
Adoption Group, we have compiled every single Java issue we have
encountered, and reported when it was new, for the latest versions of Java,
on a single page:

https://josm.openstreetmap.de/wiki/JavaBugs

We will use this page to coordinate our communication efforts with you. Is
it possible to add a link to it in the last column of the Quality Outreach
table?

The list currently contains 29 unresolved items and 18 resolved ones.

I won't go through all of them in this e-mail. Ideally we'd like to see all
of them fixed in a future Java release but I will only focus on the most
important ones.

Concerning Jigsaw:
- We have reported 3 bugs. All made it to the public JIRA: 8138878,
8140477, 8140481. The second one is a bit problematic for our tests as it
basically kills our Jenkins instance. I see the two other ones are
understood/in progress. We will do more tests after we resolve the hanging
problem.
- We'd like to know if it can be expected to see the
package sun.security.x509 become a public JDK API, for example in
javax.security.cert? We currently use it to generate a self-signed
certificate in order to create a local https server. That's our only use of
private JDK API.

Concerning Valhalla:
- I see some discussions about building the JDK with project Valhalla. Are
you going to provide public Early Access builds like project Jigsaw?

Concerning the JIRA database:
- Is it possible to add the label "josm-found" to
issues 8140481, 8139659, 8034224, 7158257, 7194099 ?
- Some issues didn't make it to the public JIRA and remained in the private
bug database. Can we please have more information on them (why have they
apparently been rejected)? The incident numbers
are JI-9009025, JI-9010791, JI-9009449, JI-9008003.
- Is it expected to allow external people to have the possibility to
subscribe to JDK issues?

Concerning our incoming migration from Java 7 to Java 8:
- I am concerned about three issues in Java2D/AWT on Linux. We have several
duplicate bug reports for them: 6322854, 7172749, 8098530. Can we hope for
a fix in a future update of Java 8?

Finally:
- We had a terrible experience when trying to report a bug against JAXP. We
detected a severe data corruption problem in StaX when dealing with Unicode
SMP characters, so we reported it, including a sample Java program 100%
reproducible, in January 2013 (JAXP-76 on java.net JIRA). As no activity
was visible on this JIRA instance, we tried to use the standard Java bug
report, three times, without success, with incident numbers 2431783
(2013-01-23), 2627098 (2013-10-28) and 9048481 (2014-11-28), without any
answer. On 2014, November 29th we discovered by chance that the bug had
finally been detected and fixed internally, as JDK-8058175 (created and
resolved in September 2014). We reported back to the public JAXP JIRA
instance, again without any answer. 6 months later we finally got the
ironic and laconic answer "Please report issues to the OpenJDK Bug System",
which was exactly was we were trying to do for 2 years! Can you please tell
us why our bug reports were all silently ignored while the bug was real,
and if is it still worth reporting bugs against JAXP? Thankfully we had far
better experiences with other components of the JDK.

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Rory O'Donnell]

On 30/10/2015 15:29, Vincent Privat wrote:

Hi,
Following the recent inclusion of JOSM to the Quality Outreach list by 
the Adoption Group, we have compiled every single Java issue we have 
encountered, and reported when it was new, for the latest versions of 
Java, on a single page:


https://josm.openstreetmap.de/wiki/JavaBugs

We will use this page to coordinate our communication efforts with 
you. Is it possible to add a link to it in the last column of the 
Quality Outreach table?
Added to the wiki, thanks for logging all these issues. I suggest you 
followup with the
relevant mailing lists to get specific feedback as you have done with 
jigsaw-dev.




The list currently contains 29 unresolved items and 18 resolved ones.

I won't go through all of them in this e-mail. Ideally we'd like to 
see all of them fixed in a future Java release but I will only focus 
on the most important ones.


Concerning Jigsaw:
- We have reported 3 bugs. All made it to the public JIRA: 8138878, 
8140477, 8140481. The second one is a bit problematic for our tests as 
it basically kills our Jenkins instance. I see the two other ones are 
understood/in progress. We will do more tests after we resolve the 
hanging problem.
- We'd like to know if it can be expected to see the 
package sun.security.x509 become a public JDK API, for example in 
javax.security.cert? We currently use it to generate a self-signed 
certificate in order to create a local https server. That's our only 
use of private JDK API.


Concerning Valhalla:
- I see some discussions about building the JDK with project Valhalla. 
Are you going to provide public Early Access builds like project Jigsaw?


Concerning the JIRA database:
- Is it possible to add the label "josm-found" to 
issues 8140481, 8139659, 8034224, 7158257, 7194099 ?
- Some issues didn't make it to the public JIRA and remained in the 
private bug database. Can we please have more information on them (why 
have they apparently been rejected)? The incident numbers 
are JI-9009025, JI-9010791, JI-9009449, JI-9008003.
- Is it expected to allow external people to have the possibility to 
subscribe to JDK issues?

I will take a look at these and come back to you.


Concerning our incoming migration from Java 7 to Java 8:
- I am concerned about three issues in Java2D/AWT on Linux. We have 
several duplicate bug reports for them: 6322854, 7172749, 8098530. Can 
we hope for a fix in a future update of Java 8?


Finally:
- We had a terrible experience when trying to report a bug against 
JAXP. We detected a severe data corruption problem in StaX when 
dealing with Unicode SMP characters, so we reported it, including a 
sample Java program 100% reproducible, in January 2013 (JAXP-76 on 
java.net  JIRA). As no activity was visible on this 
JIRA instance, we tried to use the standard Java bug report, three 
times, without success, with incident numbers 2431783 
(2013-01-23), 2627098 (2013-10-28) and 9048481 (2014-11-28), without 
any answer. On 2014, November 29th we discovered by chance that the 
bug had finally been detected and fixed internally, as JDK-8058175 
(created and resolved in September 2014). We reported back to the 
public JAXP JIRA instance, again without any answer. 6 months later we 
finally got the ironic and laconic answer "Please report issues to the 
OpenJDK Bug System", which was exactly was we were trying to do for 2 
years! Can you please tell us why our bug reports were all silently 
ignored while the bug was real, and if is it still worth reporting 
bugs against JAXP? Thankfully we had far better experiences with other 
components of the JDK.


I will take a look for you and see what I can find.

Rgds,Rory

--
Rgds,Rory O'Donnell
Quality Engineering Manager
Oracle EMEA , Dublin, Ireland

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Mandy Chung]

Thanks for reporting these issues.

> On Oct 30, 2015, at 8:29 AM, Vincent Privat  
> wrote:
> 
> Concerning Jigsaw:
> - We have reported 3 bugs. All made it to the public JIRA: 8138878,
> 8140477, 8140481. The second one is a bit problematic for our tests as it
> basically kills our Jenkins instance. I see the two other ones are
> understood/in progress. We will do more tests after we resolve the hanging
> problem.

8140481 is a build issue that should be fixed shortly.

8138878 breaks through resource encapsulation.  jake b86 fixes InternalError 
and throws MissingResourceException if it attempts to get a bundle in 
java.desktop named module.

We’ll look into these bugs.

> 
> - We'd like to know if it can be expected to see the
> package sun.security.x509 become a public JDK API, for example in
> javax.security.cert? We currently use it to generate a self-signed
> certificate in order to create a local https server. That's our only use of
> private JDK API.

There are two RFEs related to signing and certificates
  8058778: New APIs for some keytool functions
  8056174: New APIs for jar signing

I have added this comment in 8058778 for the security team to look into.  You 
can subscribe to security-...@openjdk.java.net  where the discussion for these 
RFEs will be.

Mandy

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Alan Bateman]

On 30/10/2015 15:29, Vincent Privat wrote:

:

Concerning Jigsaw:
- We have reported 3 bugs. All made it to the public JIRA: 8138878, 
8140477, 8140481. The second one is a bit problematic for our tests as 
it basically kills our Jenkins instance. I see the two other ones are 
understood/in progress. We will do more tests after we resolve the 
hanging problem.
JDK-8138878 seems to be JSOM hacking into private field in 
java.awt.Toolkit. It is possible to discuss the issue on awt-dev and 
i18n-dev to see if the changing locale scenario is meant to work?


We'll look into JDK-8140477. Note that there has been deadlocks 
involving instrumentation going back to JDK 5 (when java.lang.instrument 
was introduced). I don't know if this one is in that bucket but 
generally agents need to be take great care when instrumenting code 
defined to the ext and boot loader (because the agent's classes are 
defined by the app class loader).


JDK-8140481 seem to be the jdk.management module missing from the JDK 9 
build, it may not be specific to the Jigsaw EA builds.





:

Concerning Valhalla:
- I see some discussions about building the JDK with project Valhalla. 
Are you going to provide public Early Access builds like project Jigsaw?



This is something to ask on valhalla-dev, it might be too early for that.

-Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Jaroslav Bachorik]

On 30.10.2015 18:02, Alan Bateman wrote:



On 30/10/2015 15:29, Vincent Privat wrote:
> :
>
> Concerning Jigsaw:
> - We have reported 3 bugs. All made it to the public JIRA: 8138878,
> 8140477, 8140481. The second one is a bit problematic for our tests as
> it basically kills our Jenkins instance. I see the two other ones are
> understood/in progress. We will do more tests after we resolve the
> hanging problem.
JDK-8138878 seems to be JSOM hacking into private field in
java.awt.Toolkit. It is possible to discuss the issue on awt-dev and
i18n-dev to see if the changing locale scenario is meant to work?

We'll look into JDK-8140477. Note that there has been deadlocks
involving instrumentation going back to JDK 5 (when java.lang.instrument
was introduced). I don't know if this one is in that bucket but
generally agents need to be take great care when instrumenting code
defined to the ext and boot loader (because the agent's classes are
defined by the app class loader).

JDK-8140481 seem to be the jdk.management module missing from the JDK 9
build, it may not be specific to the Jigsaw EA builds.


No, it's not. Same problem with non-jigsaw builds.

-JB-





> :
>
> Concerning Valhalla:
> - I see some discussions about building the JDK with project Valhalla.
> Are you going to provide public Early Access builds like project Jigsaw?
>
This is something to ask on valhalla-dev, it might be too early for that.

-Alan

Re: JOSM feedback on Java 7,8,9, including Jigsaw EA

[Vincent Privat]

2015-10-30 18:02 GMT+01:00 Alan Bateman :

>
> JDK-8138878 seems to be JSOM hacking into private field in
> java.awt.Toolkit. It is possible to discuss the issue on awt-dev and
> i18n-dev to see if the changing locale scenario is meant to work?
>
>
Yes, this is a hack we found in the Sun bug database as a workaround
to JDK-6292739.
The bug describes the changing locale scenario:

"If you change the default locale at runtime, the text displayed for
accelerator keys doesn't change, e.g. if you start with default locale
Locale.GERMAN an accelerator for "control I", which shows up as "Strg-I",
doesn't change to "Ctrl-I" when changing the Locale to Locale.US."

As JDK-6292739 is classified to "Bug" we assume it is meant to work. If
this bug is resolved we can remove the hack from JOSM.