Prevent FaceletContext.FACELET_CONTEXT_KEY constant to be inlined by compiler

[glassfishrobot]

Affected Versions

[2.3]

[glassfishrobot]

Reported by lfryc

[glassfishrobot]

Reported by lfryc

[glassfishrobot]

Issue-Links:
depends on
JAVASERVERFACES-3372

[glassfishrobot]

Issue-Links:
depends on
JAVASERVERFACES-3372

[glassfishrobot]

lfryc said:
I can't see the description of the issue even though I added it.

---

Re-posting:

In RichFaces 5 when we compile ActionListenerHandler against Mojarra 2.2.5 and then run the code against older 2.1 release (2.1.7), we get NullPointerException.

The issue is well described in RF-13472.

The problem here is that a value of FACELET_CONTEXT_KEY contstant is inlined by compiler, but the value has changed between Mojarra 2.1 ("com.sun.faces.facelets.FACELET_CONTEXT") and 2.2 ("javax.faces.FACELET_CONTEXT"). This means the code compiled against one 2.2 won't work on 2.1 and vice versa.

---

What we can do is prevent compiler from inlining the FACELET_CONTEXT_KEY constant.

As suggested on stackoverflow, it is possible to use String#intern() or String#toString():

public static final String FACELET_CONTEXT_KEY = 
            "javax.faces.FACELET_CONTEXT".intern();

http://stackoverflow.com/questions/377819/are-all-compile-time-constants-inlined
http://stackoverflow.com/questions/1833581/when-to-use-intern-on-string-literals

---

I believe the fix should go to 2.1.x branch as well.

[glassfishrobot]

lfryc said:
I can't see the description of the issue even though I added it.

---

Re-posting:

In RichFaces 5 when we compile ActionListenerHandler against Mojarra 2.2.5 and then run the code against older 2.1 release (2.1.7), we get NullPointerException.

The issue is well described in RF-13472.

The problem here is that a value of FACELET_CONTEXT_KEY contstant is inlined by compiler, but the value has changed between Mojarra 2.1 ("com.sun.faces.facelets.FACELET_CONTEXT") and 2.2 ("javax.faces.FACELET_CONTEXT"). This means the code compiled against one 2.2 won't work on 2.1 and vice versa.

---

What we can do is prevent compiler from inlining the FACELET_CONTEXT_KEY constant.

As suggested on stackoverflow, it is possible to use String#intern() or String#toString():

public static final String FACELET_CONTEXT_KEY = 
            "javax.faces.FACELET_CONTEXT".intern();

http://stackoverflow.com/questions/377819/are-all-compile-time-constants-inlined
http://stackoverflow.com/questions/1833581/when-to-use-intern-on-string-literals

---

I believe the fix should go to 2.1.x branch as well.

[glassfishrobot]

File: changebundle.txt
Attached By: @manfredriem

[glassfishrobot]

File: changebundle.txt
Attached By: @manfredriem

[glassfishrobot]

@manfredriem said:
Lukas, I have attached the change bundle and once review has done I will commit it, then when it passes the 2.2 TCK I'll will close this issue. For the 2.1 work please file a back port task pointing back to this issue. Thanks!

[glassfishrobot]

@manfredriem said:
Lukas, I have attached the change bundle and once review has done I will commit it, then when it passes the 2.2 TCK I'll will close this issue. For the 2.1 work please file a back port task pointing back to this issue. Thanks!

[glassfishrobot]

@manfredriem said:
Applied to 2.2 branch,

svn commit -m "Fixes https://java.net/jira/browse/JAVASERVERFACES-3155, r=edburns, intern the FACELET_CONTEXT_KEY so it does not get inlined."
Sending jsf-api/src/main/java/javax/faces/view/facelets/FaceletContext.java
Transmitting file data .
Committed revision 12799.

Once the 2.2 TCK passes I will go ahead and close the issue. Thanks!

[glassfishrobot]

@manfredriem said:
Applied to 2.2 branch,

svn commit -m "Fixes https://java.net/jira/browse/JAVASERVERFACES-3155, r=edburns, intern the FACELET_CONTEXT_KEY so it does not get inlined."
Sending jsf-api/src/main/java/javax/faces/view/facelets/FaceletContext.java
Transmitting file data .
Committed revision 12799.

Once the 2.2 TCK passes I will go ahead and close the issue. Thanks!

[glassfishrobot]

@manfredriem said:
Applied to 2.2 branch,

svn commit -m "Reverting change as it breaks the TCK"
Sending jsf-api/src/main/java/javax/faces/view/facelets/FaceletContext.java
Transmitting file data .
Committed revision 12804.

[glassfishrobot]

@manfredriem said:
Applied to 2.2 branch,

svn commit -m "Reverting change as it breaks the TCK"
Sending jsf-api/src/main/java/javax/faces/view/facelets/FaceletContext.java
Transmitting file data .
Committed revision 12804.

[glassfishrobot]

@manfredriem said:
Unfortunately as the proposed change breaks the TCK I have moved it to the spec issue tracker to put it on the table for 2.3.

[glassfishrobot]

@manfredriem said:
Unfortunately as the proposed change breaks the TCK I have moved it to the spec issue tracker to put it on the table for 2.3.

[glassfishrobot]

@edburns said:
Set priority to baseline ahead of JSF 2.3 triage. Priorities will be assigned accurately after this exercise.

[glassfishrobot]

@edburns said:
Set priority to baseline ahead of JSF 2.3 triage. Priorities will be assigned accurately after this exercise.

[glassfishrobot]

@manfredriem said:
Setting priority to Critical

[glassfishrobot]

@manfredriem said:
Setting priority to Critical

[glassfishrobot]

Marked as fixed on Tuesday, August 19th 2014, 8:27:55 am

[glassfishrobot]

Marked as fixed on Tuesday, August 19th 2014, 8:27:55 am

[glassfishrobot]

This issue was imported from java.net JIRA JAVASERVERFACES_SPEC_PUBLIC-1257

[glassfishrobot]

This issue was imported from java.net JIRA JAVASERVERFACES_SPEC_PUBLIC-1257