I do believe in making things right (like killing Unsafe, hiding private APIs).
The transition will just be so much pain now that there is so much accumulated private APIs used in production. If there were no accessible private APIs to start with, there would be more pressure to fix bugs and expose functionality in public APIs. And no, in practice, there are not /*years*/ to transition. In practice, projects have dependencies. If I decide to move on to 9, but my dependency has not been ported yet, I have a problem. If I'm still stuck with 8, but my dependency has moved on to 9 and stopped backporting bugfixes and new features to 8 (which would be understandable, because it is quite a bit of maintanence cost to maintain two codebases), then I have a problem. Anyway, the transition is inevitable and good in the long term. I will just be happy if Oracle manages to flesh out their experimental private APIs that people are already using today, but I am somewhat skeptical that this will happen for JavaFX before 9, thus resulting in a period of temporarily lost functionality. Tomas On Wed, Apr 8, 2015 at 3:15 PM, Donald Smith <donald.sm...@oracle.com> wrote: > Making any theoretical flag available to the deployment side would entirely > miss the point. > > Let me be blunt -- sun.misc.Unsafe must die in a fire. It is -- wait for it > -- Unsafe. It must go. Ignore any kind of theoretical rope and start the > path to righteousness _*/now/*_. It is still years until the end of public > updates to JDK 8, so we have /*years */to work this out properly. But > sticking our heads in the collective sands and hoping for trivial work > arounds to Unsafe is not going to work. If you're using Unsafe, this is the > year to explain where the API is broken and get it straight.... > > Please help us kill Unsafe, kill Unsafe dead, kill Unsafe right, and do so > as quickly as possible to the ultimate benefit of everyone. > > - Don > > > On 08/04/2015 2:56 PM, Stefan Fuchs wrote: >> >> Hi, >> >> then I can only hope, that this flag is available to webstart >> applications. >> Webstart applications have no control over the installed jre. In the past >> we encountered various bugs in the jre, which required using internal apis >> for workarounds. >> For example in some releases of Java 7 the swing gui thread did not start >> unless hacking internal apis (see >> https://javafx-jira.kenai.com/browse/RT-31205 for details). If such an error >> occurs again in the future and we are no longer able to hack around the >> problem, our only choice to keep our business alive, is to discourage users >> from upgrading to newer versions of the jre, exposing them to security >> risks. >> >> - Stefan >> >> >>> >>> > it's not strictly JFX-only. >>> >>> Its not remotely FX only, in fact I could argue FX is not so affected, >>> as being relatively new it does not have 20 years of accumulation >>> of people using internal APIs that the larger JDK does, often dating from >>> when there were no suitable public APIs. There still remains some >>> of that with sun.misc.Unsafe as pointed out which will indeed be >>> inaccessible in modular mode. But the FX list isn't really the place >>> for that discussion. The jigsaw-dev is the appropriate list. FX >>> is simply bound by the rules that are set there. >>> >>> There will be a -XX flag in JDK 9 that jigsaw provides to aid in the >>> transition. >>> >>> Also remember FX is open source. You can propose patches ! >>> If there are specific APIs that are missing from FX that are suitable >>> to be *supported* public APIs then those could be considered here (this >>> list). >>> >>> -phil. >>> >>> On 4/8/2015 9:28 AM, Mike Hearn wrote: >>>> >>>> sed -i 's/private/public/g' ;) >>>> >>>> The whole notion of a strongly enforced private keyword is IMHO dumb >>>> when >>>> not using sandboxing. The number of gross hacks that occur in an attempt >>>> to >>>> work around overly strict enforcement of this stuff is crazy. The D >>>> compiler has a special flag that disables visibility enforcement when >>>> compiling unit tests, and that's a good idea, but why not go all the way >>>> and just make accessing of private state a compiler warning a la >>>> deprecated? >>>> >>>> I also need to use private JFX APIs. I think any real JFX app does, way >>>> too >>>> much basic stuff relies on it. Heck, the number of popular Java >>>> libraries >>>> that depend on sun.misc.Unsafe is huge. If Java 9 stabs us in the back >>>> in >>>> this regard then I will just write a simple tool that flips >>>> private->public >>>> either at the source level or via bytecode editing, and see what happens >>>> :-) >>>> >>>> >>>> On Wed, Apr 8, 2015 at 6:14 PM, Robert Krüger <krue...@lesspain.de> >>>> wrote: >>>> >>>>> Hi, >>>>> >>>>> I hope this is not too off-topic, because although it came up in a JFX >>>>> context it's not strictly JFX-only. >>>>> >>>>> Someone from our team recently had a chat with a high-ranking regional >>>>> Oracle representative who gave a talk on the state of JFX. Our guy >>>>> explained our situation (evaluating JFX to migrate our swing-based >>>>> product, >>>>> feeling it's in principle the right technology but still having >>>>> show-stopping limitations like RT-36215) and the Oracle guy offered to >>>>> relay our concrete questions to the right people, which he did. >>>>> >>>>> The answer we got contained one thing that really was a bit of a shock >>>>> and >>>>> I would like someone to either confirm this or clear up a >>>>> misunderstanding. >>>>> >>>>> The statement was that private APIs will not be available in JDK 9 due >>>>> to >>>>> modularity restrictions. If that is the case and we no longer have the >>>>> ability to build temporary workarounds using private APIs (which in our >>>>> case is controllable as we ship the JRE with our product), I would >>>>> probably >>>>> have to stop any development going into the direction of JFX as we will >>>>> probably have to use 9 at some point because many things now scheduled >>>>> for >>>>> 9 will not get fixed in 8 and we will most likely still need >>>>> workarounds >>>>> using private API, at least that's what my current experience with JFX >>>>> tells me. >>>>> >>>>> Please tell me that this was a misunderstanding (maybe meant for the >>>>> general case where one does not ship the JRE) or a non-engineering >>>>> source >>>>> that simply made mistake. >>>>> >>>>> Best regards and thanks in advance, >>>>> >>>>> Robert >>>>> >>> >>> >> >
