You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openjpa.apache.org by Mark Struberg <st...@yahoo.de.INVALID> on 2016/09/15 13:49:21 UTC
state of OpenJPA on Java9
Hi folks!
I did invest some time to get OpenJPA running on Java9.
We did hit some issues, mainly with the stricter behaviour of getResources with Jigsaw.
In our enhancement layer we pretty often load just the byte[] of Classes to avoid triggering the actual Classloading. This is mostly done by
ClassLoader#getResourceAsStream(className.replace('.', '/') + ".class");
But this just returns null in Java9.
This currently breaks a LOT of apps, so we are far from being the only one.
After talking with Oracle and a few other OSS people, Mark Reinhold agreed to address this issue:
http://mail.openjdk.java.net/pipermail/jpms-spec-experts/2016-September/000392.html
A revised Java9-EA should be available at the end of this week.
If OpenJPA works with that version, then there is also no urgent need to get rid of Serp.
In which case I'd be good to to an openjpa-3.0.0-alpha-1 release. Wdyt?
LieGrue,
strub
Re: state of OpenJPA on Java9
Posted by Romain Manni-Bucau <rm...@gmail.com>.
Le 15 sept. 2016 21:17, "Mark Struberg" <st...@yahoo.de.invalid> a
écrit :
>
> I'd rather ship 3.0.x now and keep working on the ASM move in the branch.
>
> And then enable it either pluggable or move the default to ASM for 3.1.x.
> It's really much more work than you think...
>
Well it will be a big breaking - whatever we do - change so should be
another major but np with 4.x.
That said no reason to use 3.x if we are not yet 100% targetting jpa 2.1
too so 2.5 was a compromise.
> LieGrue,
> strub
>
>
>
>
>
> > On Thursday, 15 September 2016, 21:01, Romain Manni-Bucau <
rmannibucau@gmail.com> wrote:
> > > Being said i cant help much ATM (:'() not sure i can judge but making
it
> > part of 3.0 would be important IMO and qane for the future. Otherwise
what
> > about doing some 2.5 with stored proc and std ddl for instance?
> >
> > Le 15 sept. 2016 20:47, "Mark Struberg"
> > <st...@yahoo.de.invalid> a
> > écrit :
> >
> >
> >> Of course it would be better to have the bytecode stuff pluggable and
move
> >> to ASM for default.
> >>
> >> The problem is that there are quite a few classes which are heavily
bound
> >> to Serp by using the BCClass interface as parameter type. E.g.
> >> AuxilaryEnhancer.
> >>
> >> The generated bytecode is also very complex (might get streamlined
ofc).
> >>
> >> The question is really: IF Serp works on Java9, is it a high prio for
> >> 3.0.0 to get rid of Serp? Or is moving over to ASM just a long-time
goal in
> >> that case?
> >>
> >> I really would love to ship a 3.0.0 release (probably beta), so
people can
> >> start using JPA-2.1 with OpenJPA.
> >>
> >>
> >> LieGrue,
> >> strub
> >>
> >>
> >>
> >>
> >>
> >> > On Thursday, 15 September 2016, 20:37, Romain Manni-Bucau <
> >> rmannibucau@gmail.com> wrote:
> >> > > Since the jvm uses asm - shaded - i dee this convergence as only
> > being
> >> > positive for our ecosystem if reachable.
> >> >
> >> > Le 15 sept. 2016 20:34, "Mark Struberg"
> >> > <st...@yahoo.de.invalid> a
> >> > écrit :
> >> >
> >> >
> >> >> Yes, most likely it could be seen as regression.
> >> >>
> >> >> The problem was that it was not easy to spot as Serp was blowing
> > up
> >> wildly.
> >> >>
> >> >> Serp has quite a few empty catch(Exception) blocks :(
> >> >> In conjunction with the heavy use of magic indices + bit flags I
> > simply
> >> >> only saw random side effects until I discovered the real reason
> > (took
> >> me 2
> >> >> weeks).
> >> >>
> >> >> LieGrue,
> >> >> strub
> >> >>
> >> >>
> >> >>
> >> >> On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <
> >> >> rmannibucau@gmail.com> wrote:
> >> >> >This issue hits most of projects including a plain tomcat so
> > guess j9
> >> >> >should avoid this regression (yes it is).
> >> >> >
> >> >> >Le 15 sept. 2016 16:14, "Francesco Chicchiriccò"
> >> > <il...@apache.org> a
> >> >> >écrit :
> >> >> >
> >> >> >
> >> >> >> On 15/09/2016 15:49, Mark Struberg wrote:
> >> >> >>
> >> >> >>> Hi folks!
> >> >> >>>
> >> >> >>> I did invest some time to get OpenJPA running on
> > Java9.
> >> >> >>> We did hit some issues, mainly with the stricter
> > behaviour of
> >> >> >>> getResources with Jigsaw.
> >> >> >>>
> >> >> >>> In our enhancement layer we pretty often load just
> > the byte[]
> >> > of
> >> >> Classes
> >> >> >>> to avoid triggering the actual Classloading. This is
> > mostly
> >> > done by
> >> >> >>>
> >> >> >>>
> > ClassLoader#getResourceAsStream(className.replace('.',
> >> > '/') +
> >> >> ".class");
> >> >> >>> But this just returns null in Java9.
> >> >> >>>
> >> >> >>>
> >> >> >>> This currently breaks a LOT of apps, so we are far
> > from being
> >> > the only
> >> >> >>> one.
> >> >> >>> After talking with Oracle and a few other OSS
> > people, Mark
> >> > Reinhold
> >> >> >>> agreed to address this issue:
> >> >> >>>
> > http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
> >> >> >>> 6-September/000392.html
> >> >> >>>
> >> >> >>> A revised Java9-EA should be available at the end of
> > this
> >> > week.
> >> >> >>> If OpenJPA works with that version, then there is
> > also no
> >> > urgent need
> >> >> to
> >> >> >>> get rid of Serp.
> >> >> >>>
> >> >> >>> In which case I'd be good to to an
> > openjpa-3.0.0-alpha-1
> >> > release. Wdyt?
> >> >> >>>
> >> >> >>
> >> >> >>
> >> >> >> +1!
> >> >> >> Regards.
> >> >> >>
> >> >> >> --
> >> >> >> Francesco Chicchiriccò
> >> >> >>
> >> >> >> Tirasa - Open Source Excellence
> >> >> >> http://www.tirasa.net/
> >> >> >>
> >> >> >> Involved at The Apache Software Foundation:
> >> >> >> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
> >> >> >> CXF Committer, OpenJPA Committer, PonyMail PPMC
> >> >> >> http://home.apache.org/~ilgrosso/
> >> >> >>
> >> >> >>
> >> >> >
> >> >> >
> >> >>
> >> >
> >>
> >
Re: state of OpenJPA on Java9
Posted by Mark Struberg <st...@yahoo.de.INVALID>.
I'd rather ship 3.0.x now and keep working on the ASM move in the branch.
And then enable it either pluggable or move the default to ASM for 3.1.x.
It's really much more work than you think...
LieGrue,
strub
> On Thursday, 15 September 2016, 21:01, Romain Manni-Bucau <rm...@gmail.com> wrote:
> > Being said i cant help much ATM (:'() not sure i can judge but making it
> part of 3.0 would be important IMO and qane for the future. Otherwise what
> about doing some 2.5 with stored proc and std ddl for instance?
>
> Le 15 sept. 2016 20:47, "Mark Struberg"
> <st...@yahoo.de.invalid> a
> écrit :
>
>
>> Of course it would be better to have the bytecode stuff pluggable and move
>> to ASM for default.
>>
>> The problem is that there are quite a few classes which are heavily bound
>> to Serp by using the BCClass interface as parameter type. E.g.
>> AuxilaryEnhancer.
>>
>> The generated bytecode is also very complex (might get streamlined ofc).
>>
>> The question is really: IF Serp works on Java9, is it a high prio for
>> 3.0.0 to get rid of Serp? Or is moving over to ASM just a long-time goal in
>> that case?
>>
>> I really would love to ship a 3.0.0 release (probably beta), so people can
>> start using JPA-2.1 with OpenJPA.
>>
>>
>> LieGrue,
>> strub
>>
>>
>>
>>
>>
>> > On Thursday, 15 September 2016, 20:37, Romain Manni-Bucau <
>> rmannibucau@gmail.com> wrote:
>> > > Since the jvm uses asm - shaded - i dee this convergence as only
> being
>> > positive for our ecosystem if reachable.
>> >
>> > Le 15 sept. 2016 20:34, "Mark Struberg"
>> > <st...@yahoo.de.invalid> a
>> > écrit :
>> >
>> >
>> >> Yes, most likely it could be seen as regression.
>> >>
>> >> The problem was that it was not easy to spot as Serp was blowing
> up
>> wildly.
>> >>
>> >> Serp has quite a few empty catch(Exception) blocks :(
>> >> In conjunction with the heavy use of magic indices + bit flags I
> simply
>> >> only saw random side effects until I discovered the real reason
> (took
>> me 2
>> >> weeks).
>> >>
>> >> LieGrue,
>> >> strub
>> >>
>> >>
>> >>
>> >> On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <
>> >> rmannibucau@gmail.com> wrote:
>> >> >This issue hits most of projects including a plain tomcat so
> guess j9
>> >> >should avoid this regression (yes it is).
>> >> >
>> >> >Le 15 sept. 2016 16:14, "Francesco Chicchiriccò"
>> > <il...@apache.org> a
>> >> >écrit :
>> >> >
>> >> >
>> >> >> On 15/09/2016 15:49, Mark Struberg wrote:
>> >> >>
>> >> >>> Hi folks!
>> >> >>>
>> >> >>> I did invest some time to get OpenJPA running on
> Java9.
>> >> >>> We did hit some issues, mainly with the stricter
> behaviour of
>> >> >>> getResources with Jigsaw.
>> >> >>>
>> >> >>> In our enhancement layer we pretty often load just
> the byte[]
>> > of
>> >> Classes
>> >> >>> to avoid triggering the actual Classloading. This is
> mostly
>> > done by
>> >> >>>
>> >> >>>
> ClassLoader#getResourceAsStream(className.replace('.',
>> > '/') +
>> >> ".class");
>> >> >>> But this just returns null in Java9.
>> >> >>>
>> >> >>>
>> >> >>> This currently breaks a LOT of apps, so we are far
> from being
>> > the only
>> >> >>> one.
>> >> >>> After talking with Oracle and a few other OSS
> people, Mark
>> > Reinhold
>> >> >>> agreed to address this issue:
>> >> >>>
> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
>> >> >>> 6-September/000392.html
>> >> >>>
>> >> >>> A revised Java9-EA should be available at the end of
> this
>> > week.
>> >> >>> If OpenJPA works with that version, then there is
> also no
>> > urgent need
>> >> to
>> >> >>> get rid of Serp.
>> >> >>>
>> >> >>> In which case I'd be good to to an
> openjpa-3.0.0-alpha-1
>> > release. Wdyt?
>> >> >>>
>> >> >>
>> >> >>
>> >> >> +1!
>> >> >> Regards.
>> >> >>
>> >> >> --
>> >> >> Francesco Chicchiriccò
>> >> >>
>> >> >> Tirasa - Open Source Excellence
>> >> >> http://www.tirasa.net/
>> >> >>
>> >> >> Involved at The Apache Software Foundation:
>> >> >> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
>> >> >> CXF Committer, OpenJPA Committer, PonyMail PPMC
>> >> >> http://home.apache.org/~ilgrosso/
>> >> >>
>> >> >>
>> >> >
>> >> >
>> >>
>> >
>>
>
Re: state of OpenJPA on Java9
Posted by Romain Manni-Bucau <rm...@gmail.com>.
Being said i cant help much ATM (:'() not sure i can judge but making it
part of 3.0 would be important IMO and qane for the future. Otherwise what
about doing some 2.5 with stored proc and std ddl for instance?
Le 15 sept. 2016 20:47, "Mark Struberg" <st...@yahoo.de.invalid> a
écrit :
> Of course it would be better to have the bytecode stuff pluggable and move
> to ASM for default.
>
> The problem is that there are quite a few classes which are heavily bound
> to Serp by using the BCClass interface as parameter type. E.g.
> AuxilaryEnhancer.
>
> The generated bytecode is also very complex (might get streamlined ofc).
>
> The question is really: IF Serp works on Java9, is it a high prio for
> 3.0.0 to get rid of Serp? Or is moving over to ASM just a long-time goal in
> that case?
>
> I really would love to ship a 3.0.0 release (probably beta), so people can
> start using JPA-2.1 with OpenJPA.
>
>
> LieGrue,
> strub
>
>
>
>
>
> > On Thursday, 15 September 2016, 20:37, Romain Manni-Bucau <
> rmannibucau@gmail.com> wrote:
> > > Since the jvm uses asm - shaded - i dee this convergence as only being
> > positive for our ecosystem if reachable.
> >
> > Le 15 sept. 2016 20:34, "Mark Struberg"
> > <st...@yahoo.de.invalid> a
> > écrit :
> >
> >
> >> Yes, most likely it could be seen as regression.
> >>
> >> The problem was that it was not easy to spot as Serp was blowing up
> wildly.
> >>
> >> Serp has quite a few empty catch(Exception) blocks :(
> >> In conjunction with the heavy use of magic indices + bit flags I simply
> >> only saw random side effects until I discovered the real reason (took
> me 2
> >> weeks).
> >>
> >> LieGrue,
> >> strub
> >>
> >>
> >>
> >> On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <
> >> rmannibucau@gmail.com> wrote:
> >> >This issue hits most of projects including a plain tomcat so guess j9
> >> >should avoid this regression (yes it is).
> >> >
> >> >Le 15 sept. 2016 16:14, "Francesco Chicchiriccò"
> > <il...@apache.org> a
> >> >écrit :
> >> >
> >> >
> >> >> On 15/09/2016 15:49, Mark Struberg wrote:
> >> >>
> >> >>> Hi folks!
> >> >>>
> >> >>> I did invest some time to get OpenJPA running on Java9.
> >> >>> We did hit some issues, mainly with the stricter behaviour of
> >> >>> getResources with Jigsaw.
> >> >>>
> >> >>> In our enhancement layer we pretty often load just the byte[]
> > of
> >> Classes
> >> >>> to avoid triggering the actual Classloading. This is mostly
> > done by
> >> >>>
> >> >>> ClassLoader#getResourceAsStream(className.replace('.',
> > '/') +
> >> ".class");
> >> >>> But this just returns null in Java9.
> >> >>>
> >> >>>
> >> >>> This currently breaks a LOT of apps, so we are far from being
> > the only
> >> >>> one.
> >> >>> After talking with Oracle and a few other OSS people, Mark
> > Reinhold
> >> >>> agreed to address this issue:
> >> >>> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
> >> >>> 6-September/000392.html
> >> >>>
> >> >>> A revised Java9-EA should be available at the end of this
> > week.
> >> >>> If OpenJPA works with that version, then there is also no
> > urgent need
> >> to
> >> >>> get rid of Serp.
> >> >>>
> >> >>> In which case I'd be good to to an openjpa-3.0.0-alpha-1
> > release. Wdyt?
> >> >>>
> >> >>
> >> >>
> >> >> +1!
> >> >> Regards.
> >> >>
> >> >> --
> >> >> Francesco Chicchiriccò
> >> >>
> >> >> Tirasa - Open Source Excellence
> >> >> http://www.tirasa.net/
> >> >>
> >> >> Involved at The Apache Software Foundation:
> >> >> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
> >> >> CXF Committer, OpenJPA Committer, PonyMail PPMC
> >> >> http://home.apache.org/~ilgrosso/
> >> >>
> >> >>
> >> >
> >> >
> >>
> >
>
Re: state of OpenJPA on Java9
Posted by Mark Struberg <st...@yahoo.de.INVALID>.
Of course it would be better to have the bytecode stuff pluggable and move to ASM for default.
The problem is that there are quite a few classes which are heavily bound to Serp by using the BCClass interface as parameter type. E.g. AuxilaryEnhancer.
The generated bytecode is also very complex (might get streamlined ofc).
The question is really: IF Serp works on Java9, is it a high prio for 3.0.0 to get rid of Serp? Or is moving over to ASM just a long-time goal in that case?
I really would love to ship a 3.0.0 release (probably beta), so people can start using JPA-2.1 with OpenJPA.
LieGrue,
strub
> On Thursday, 15 September 2016, 20:37, Romain Manni-Bucau <rm...@gmail.com> wrote:
> > Since the jvm uses asm - shaded - i dee this convergence as only being
> positive for our ecosystem if reachable.
>
> Le 15 sept. 2016 20:34, "Mark Struberg"
> <st...@yahoo.de.invalid> a
> écrit :
>
>
>> Yes, most likely it could be seen as regression.
>>
>> The problem was that it was not easy to spot as Serp was blowing up wildly.
>>
>> Serp has quite a few empty catch(Exception) blocks :(
>> In conjunction with the heavy use of magic indices + bit flags I simply
>> only saw random side effects until I discovered the real reason (took me 2
>> weeks).
>>
>> LieGrue,
>> strub
>>
>>
>>
>> On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <
>> rmannibucau@gmail.com> wrote:
>> >This issue hits most of projects including a plain tomcat so guess j9
>> >should avoid this regression (yes it is).
>> >
>> >Le 15 sept. 2016 16:14, "Francesco Chicchiriccò"
> <il...@apache.org> a
>> >écrit :
>> >
>> >
>> >> On 15/09/2016 15:49, Mark Struberg wrote:
>> >>
>> >>> Hi folks!
>> >>>
>> >>> I did invest some time to get OpenJPA running on Java9.
>> >>> We did hit some issues, mainly with the stricter behaviour of
>> >>> getResources with Jigsaw.
>> >>>
>> >>> In our enhancement layer we pretty often load just the byte[]
> of
>> Classes
>> >>> to avoid triggering the actual Classloading. This is mostly
> done by
>> >>>
>> >>> ClassLoader#getResourceAsStream(className.replace('.',
> '/') +
>> ".class");
>> >>> But this just returns null in Java9.
>> >>>
>> >>>
>> >>> This currently breaks a LOT of apps, so we are far from being
> the only
>> >>> one.
>> >>> After talking with Oracle and a few other OSS people, Mark
> Reinhold
>> >>> agreed to address this issue:
>> >>> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
>> >>> 6-September/000392.html
>> >>>
>> >>> A revised Java9-EA should be available at the end of this
> week.
>> >>> If OpenJPA works with that version, then there is also no
> urgent need
>> to
>> >>> get rid of Serp.
>> >>>
>> >>> In which case I'd be good to to an openjpa-3.0.0-alpha-1
> release. Wdyt?
>> >>>
>> >>
>> >>
>> >> +1!
>> >> Regards.
>> >>
>> >> --
>> >> Francesco Chicchiriccò
>> >>
>> >> Tirasa - Open Source Excellence
>> >> http://www.tirasa.net/
>> >>
>> >> Involved at The Apache Software Foundation:
>> >> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
>> >> CXF Committer, OpenJPA Committer, PonyMail PPMC
>> >> http://home.apache.org/~ilgrosso/
>> >>
>> >>
>> >
>> >
>>
>
Re: state of OpenJPA on Java9
Posted by Romain Manni-Bucau <rm...@gmail.com>.
Since the jvm uses asm - shaded - i dee this convergence as only being
positive for our ecosystem if reachable.
Le 15 sept. 2016 20:34, "Mark Struberg" <st...@yahoo.de.invalid> a
écrit :
> Yes, most likely it could be seen as regression.
>
> The problem was that it was not easy to spot as Serp was blowing up wildly.
>
> Serp has quite a few empty catch(Exception) blocks :(
> In conjunction with the heavy use of magic indices + bit flags I simply
> only saw random side effects until I discovered the real reason (took me 2
> weeks).
>
> LieGrue,
> strub
>
>
>
> On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <
> rmannibucau@gmail.com> wrote:
> >This issue hits most of projects including a plain tomcat so guess j9
> >should avoid this regression (yes it is).
> >
> >Le 15 sept. 2016 16:14, "Francesco Chicchiriccò" <il...@apache.org> a
> >écrit :
> >
> >
> >> On 15/09/2016 15:49, Mark Struberg wrote:
> >>
> >>> Hi folks!
> >>>
> >>> I did invest some time to get OpenJPA running on Java9.
> >>> We did hit some issues, mainly with the stricter behaviour of
> >>> getResources with Jigsaw.
> >>>
> >>> In our enhancement layer we pretty often load just the byte[] of
> Classes
> >>> to avoid triggering the actual Classloading. This is mostly done by
> >>>
> >>> ClassLoader#getResourceAsStream(className.replace('.', '/') +
> ".class");
> >>> But this just returns null in Java9.
> >>>
> >>>
> >>> This currently breaks a LOT of apps, so we are far from being the only
> >>> one.
> >>> After talking with Oracle and a few other OSS people, Mark Reinhold
> >>> agreed to address this issue:
> >>> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
> >>> 6-September/000392.html
> >>>
> >>> A revised Java9-EA should be available at the end of this week.
> >>> If OpenJPA works with that version, then there is also no urgent need
> to
> >>> get rid of Serp.
> >>>
> >>> In which case I'd be good to to an openjpa-3.0.0-alpha-1 release. Wdyt?
> >>>
> >>
> >>
> >> +1!
> >> Regards.
> >>
> >> --
> >> Francesco Chicchiriccò
> >>
> >> Tirasa - Open Source Excellence
> >> http://www.tirasa.net/
> >>
> >> Involved at The Apache Software Foundation:
> >> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
> >> CXF Committer, OpenJPA Committer, PonyMail PPMC
> >> http://home.apache.org/~ilgrosso/
> >>
> >>
> >
> >
>
Re: state of OpenJPA on Java9
Posted by Mark Struberg <st...@yahoo.de.INVALID>.
Yes, most likely it could be seen as regression.
The problem was that it was not easy to spot as Serp was blowing up wildly.
Serp has quite a few empty catch(Exception) blocks :(
In conjunction with the heavy use of magic indices + bit flags I simply only saw random side effects until I discovered the real reason (took me 2 weeks).
LieGrue,
strub
On Thursday, 15 September 2016, 19:03, Romain Manni-Bucau <rm...@gmail.com> wrote:
>This issue hits most of projects including a plain tomcat so guess j9
>should avoid this regression (yes it is).
>
>Le 15 sept. 2016 16:14, "Francesco Chicchiriccò" <il...@apache.org> a
>écrit :
>
>
>> On 15/09/2016 15:49, Mark Struberg wrote:
>>
>>> Hi folks!
>>>
>>> I did invest some time to get OpenJPA running on Java9.
>>> We did hit some issues, mainly with the stricter behaviour of
>>> getResources with Jigsaw.
>>>
>>> In our enhancement layer we pretty often load just the byte[] of Classes
>>> to avoid triggering the actual Classloading. This is mostly done by
>>>
>>> ClassLoader#getResourceAsStream(className.replace('.', '/') + ".class");
>>> But this just returns null in Java9.
>>>
>>>
>>> This currently breaks a LOT of apps, so we are far from being the only
>>> one.
>>> After talking with Oracle and a few other OSS people, Mark Reinhold
>>> agreed to address this issue:
>>> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
>>> 6-September/000392.html
>>>
>>> A revised Java9-EA should be available at the end of this week.
>>> If OpenJPA works with that version, then there is also no urgent need to
>>> get rid of Serp.
>>>
>>> In which case I'd be good to to an openjpa-3.0.0-alpha-1 release. Wdyt?
>>>
>>
>>
>> +1!
>> Regards.
>>
>> --
>> Francesco Chicchiriccò
>>
>> Tirasa - Open Source Excellence
>> http://www.tirasa.net/
>>
>> Involved at The Apache Software Foundation:
>> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
>> CXF Committer, OpenJPA Committer, PonyMail PPMC
>> http://home.apache.org/~ilgrosso/
>>
>>
>
>
Re: state of OpenJPA on Java9
Posted by Romain Manni-Bucau <rm...@gmail.com>.
This issue hits most of projects including a plain tomcat so guess j9
should avoid this regression (yes it is).
Le 15 sept. 2016 16:14, "Francesco Chicchiriccò" <il...@apache.org> a
écrit :
> On 15/09/2016 15:49, Mark Struberg wrote:
>
>> Hi folks!
>>
>> I did invest some time to get OpenJPA running on Java9.
>> We did hit some issues, mainly with the stricter behaviour of
>> getResources with Jigsaw.
>>
>> In our enhancement layer we pretty often load just the byte[] of Classes
>> to avoid triggering the actual Classloading. This is mostly done by
>>
>> ClassLoader#getResourceAsStream(className.replace('.', '/') + ".class");
>> But this just returns null in Java9.
>>
>>
>> This currently breaks a LOT of apps, so we are far from being the only
>> one.
>> After talking with Oracle and a few other OSS people, Mark Reinhold
>> agreed to address this issue:
>> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/201
>> 6-September/000392.html
>>
>> A revised Java9-EA should be available at the end of this week.
>> If OpenJPA works with that version, then there is also no urgent need to
>> get rid of Serp.
>>
>> In which case I'd be good to to an openjpa-3.0.0-alpha-1 release. Wdyt?
>>
>
>
> +1!
> Regards.
>
> --
> Francesco Chicchiriccò
>
> Tirasa - Open Source Excellence
> http://www.tirasa.net/
>
> Involved at The Apache Software Foundation:
> member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
> CXF Committer, OpenJPA Committer, PonyMail PPMC
> http://home.apache.org/~ilgrosso/
>
>
Re: state of OpenJPA on Java9
Posted by Francesco Chicchiriccò <il...@apache.org>.
On 15/09/2016 15:49, Mark Struberg wrote:
> Hi folks!
>
> I did invest some time to get OpenJPA running on Java9.
> We did hit some issues, mainly with the stricter behaviour of getResources with Jigsaw.
>
> In our enhancement layer we pretty often load just the byte[] of Classes to avoid triggering the actual Classloading. This is mostly done by
>
> ClassLoader#getResourceAsStream(className.replace('.', '/') + ".class");
> But this just returns null in Java9.
>
>
> This currently breaks a LOT of apps, so we are far from being the only one.
> After talking with Oracle and a few other OSS people, Mark Reinhold agreed to address this issue:
> http://mail.openjdk.java.net/pipermail/jpms-spec-experts/2016-September/000392.html
>
> A revised Java9-EA should be available at the end of this week.
> If OpenJPA works with that version, then there is also no urgent need to get rid of Serp.
>
> In which case I'd be good to to an openjpa-3.0.0-alpha-1 release. Wdyt?
+1!
Regards.
--
Francesco Chicchiricc�
Tirasa - Open Source Excellence
http://www.tirasa.net/
Involved at The Apache Software Foundation:
member, Syncope PMC chair, Cocoon PMC, Olingo PMC,
CXF Committer, OpenJPA Committer, PonyMail PPMC
http://home.apache.org/~ilgrosso/