On 5/03/2013 5:46 PM, Erik Joelsson wrote:
On 2013-03-05 03:25, David Holmes wrote:
On 5/03/2013 3:05 AM, Jim Laskey (Oracle) wrote:
Erik,
The changes to remove $ from the class file name (Nashorn) are in tl.
Thanks Jim.
We'll need a new bug filed to revert the original $$ change and the
subsequent \$$$$ -> \$$ changes. I can probably do that in conjunction
with some other cleanups in the profile-include lists.
Good to see this getting resolved. Tell me if you need any help with it.
All in hand - thanks Erik.
JDK-8009428: Revert changes to $ substitution performed as part of
nashorn integration
David
-----
/Erik
David
-----
https://jbs.oracle.com/bugs/browse/JDK-8009379
HG Updates added a comment - 2013-03-04 12:54
URL: http://hg.openjdk.java.net/jdk8/tl/nashorn/rev/fe5211fc3114
User: sundar
Date: 2013-03-04 16:49:50 +0000
-- Jim
On 2013-03-04, at 11:58 AM, Jim Laskey (Oracle)
<[email protected]> wrote:
I have a change set making its way to tl that removes this requirement.
-- Jim
On 2013-03-04, at 11:54 AM, Kelly O'Hair <[email protected]>
wrote:
It seems to me that using $ in class names when they are NOT Inner
Classes is a huge mistake.
Mark my words, this will come back to haunt us, multiple times, I
guarantee it.
-kto
On Mar 4, 2013, at 3:20 AM, Erik Joelsson wrote:
The change in ListPathsSafely was needed because in nashorn, there
are java files with $ in the class name (not inner classes!).
Thinking of it now, I can imagine my change there causing problems
for other uses of the macro. Hopefully it will all be better with
fewer $$ in the makefiles in the long run though.
/Erik
On 2013-02-26 21:29, David Holmes wrote:
Alan,
Thanks for diving onto this. I can't say I understand what
changed in detail yet but I know that there are places where I
had to jump through hoops to deal with $ in class names
appropriately. I would not be surprised if there is further
breakage if somehow this expansion mechanism has changed.
David
On 26/02/2013 11:48 PM, Alan Bateman wrote:
The build changes for Nashorn were pushed to jdk8/tl yesterday
and one
of the casualties is the profiles build.
My reading of the make file changes is that ListPathsSafely_If
(defined
in MakeBase.gmk) has changed the expansion so that secondary
expansion
is no longer required. Erik is away this week but I assume this was
intentional.
Attached is the diffs that I propose to push to jdk8/tl today to
get
profiles building again, assuming I get a reviewer.
-Alan
diff --git a/makefiles/profile-rtjar-includes.txt
b/makefiles/profile-rtjar-includes.txt
--- a/makefiles/profile-rtjar-includes.txt
+++ b/makefiles/profile-rtjar-includes.txt
@@ -349,6 +349,7 @@
com/sun/rowset/providers \
com/sun/script/javascript \
com/sun/script/util \
+ com/sun/security/auth \
com/sun/security/auth/callback \
com/sun/security/auth/login \
com/sun/security/auth/module \
@@ -448,8 +449,7 @@
sun/tracing \
sun/tracing/dtrace
-PROFILE_3_RTJAR_INCLUDE_TYPES := \
- com/sun/security/auth/*.class
+PROFILE_3_RTJAR_INCLUDE_TYPES :=
PROFILE_3_RTJAR_EXCLUDE_TYPES := \
javax/management/remote/rmi/_RMIConnectionImpl_Tie.class \
@@ -457,10 +457,10 @@
javax/management/remote/rmi/_RMIServerImpl_Tie.class \
javax/management/remote/rmi/_RMIServer_Stub.class \
com/sun/security/auth/callback/DialogCallbackHandler.class \
-
com/sun/security/auth/callback/DialogCallbackHandler\$$$$1.class \
-
com/sun/security/auth/callback/DialogCallbackHandler\$$$$2.class \
-
com/sun/security/auth/callback/DialogCallbackHandler\$$$$Action.class
\
-
com/sun/security/auth/callback/DialogCallbackHandler\$$$$ConfirmationInfo.class
+ com/sun/security/auth/callback/DialogCallbackHandler\$$1.class \
+ com/sun/security/auth/callback/DialogCallbackHandler\$$2.class \
+
com/sun/security/auth/callback/DialogCallbackHandler\$$Action.class
\
+
com/sun/security/auth/callback/DialogCallbackHandler\$$ConfirmationInfo.class
PROFILE_3_INCLUDE_METAINF_SERVICES := \
META-INF/services/javax.script.ScriptEngineFactory
