My 2 cents:
@SuppressWarnings("serial") is never a good idea. There will still be a
serialver automatically computed and you won't be able to control it.
Will the code changes go to jdk8 only? If so, I guess you should use the
value calculated from jdk7 codes. If the values on jdk6 and jdk7 are
already different, sorry, that means there is already a compatibility
problem made and you won't be able to fix it. We can now only make sure
jdk7 and jdk8 are compatible.
Ultimately you can write small test to see if serialized form from one
version can be accepted by another version.
Thanks
Max
On 08/04/2011 08:10 AM, Alexandre Boulgakov wrote:
Those were the ones I was talking about- the serialVersionUIDs I
mentioned were the ones generated by serialver.exe. The webrev doesn't
have any of the pkcs11 changes yet (including the added serialVersionUIDs).
I'll wait for Brad's or Valerie's response.
Thanks,
Sasha
On 8/3/2011 5:07 PM, Xuelei Fan wrote:
On 8/4/2011 7:52 AM, Alexandre Boulgakov wrote:
There is currently no serialVersionUID defined for these classes. Do you
mean that we cannot add one for backwards compatibility?
My answers only apply to you latest e-mail about the serialVersionUID
update in sun.security.pkcs11.P11Key.P11SecretKey and
sun.security.pkcs11.P11TlsPrfGenerator$1. I think you need to use the
current values unless you get an approval from PKCS11 owner.
We may need to consider more for these classes without serialVersionUID,
I will look at your webrev again if I can get some time today. Normally,
I think if there is a new attribute in a class, it is OK to add a new
serialVersionUID value.
Xuelei
If so, would
the best solution be to add an @SuppressWarnings("serial") annotation to
these classes?
Thanks,
Sasha
On 8/3/2011 4:49 PM, Xuelei Fan wrote:
Oops, I missed this.
I don't think we can modify serialVersionUID value for backward
compatibility.
Thanks,
Xuelei
On 8/4/2011 7:39 AM, Alexandre Boulgakov wrote:
Ping..?
-Sasha
On 7/27/2011 11:22 AM, Alexandre Boulgakov wrote:
Should I just use the newest serialVersionUID for both of them?
-Sasha
On 7/26/2011 10:31 AM, Alexandre Boulgakov wrote:
I just noticed that pkcs11 is not built on my machine (64-bit
Windows) so I missed all of the warnings there. There are two
mission
serialVersionUID warnings for classes that have had different
generated serialVersionUID's in the past.
sun.security.pkcs11.P11Key.P11SecretKey
-currently: -7828241727014329084L;
-JDK 1.5: -897881148551545872L;
sun.security.pkcs11.P11TlsPrfGenerator$1
-currently: -8090049519656411362L;
-JDK 6: -3305145912345854189L;
I'm not sure why the serialVersionUID changed for
sun.security.pkcs11.P11TlsPrfGenerator$1; the code is the same, and
the serialVersionUID for the base class javax.crypto.SecretKey
hasn't
changed.
For sun.security.pkcs11.P11Key.P11SecretKey, the code is the same,
but the base class sun.security.pkcs11.P11Key has changed.
How should I go about resolving these issues?
Thanks,
Sasha
On 7/20/2011 3:37 PM, [email protected] wrote:
On Jul 21, 2011, at 1:25 AM, Alexandre
Boulgakov<[email protected]> wrote:
This is a Netbeans warning, not generated by the compiler. The
reason is that List.isEmpty() can be more efficient for some
implementations. ArrayList.size() == 0 and ArrayList.isEmpty()
should take the same time, so it doesn't matter for allResults,
but
keyTypeList is a List argument, so any implementation could be
passed in. List.isEmpty() should never be slower than List.size()
== 0 because AbstractCollection defines isEmpty() as size() == 0.
Even if we don't get a performance improvement, it still improves
readability.
Sounds reasonable.
Thanks,
Xuelei
-Sasha
On 7/19/2011 7:32 PM, Xuelei Fan wrote:
I was looking at the updates in sun/security/ssl. Looks fine to
me.
It's fine, but I just wonder why List.isEmpty() is preferred to
(List.size() == 0). What's the compiler warning for (List.size()
== 0)?
src/share/classes/sun/security/ssl/X509KeyManagerImpl.java
- if (keyTypeList == null || keyTypeList.size() == 0) {
+ if (keyTypeList == null || keyTypeList.isEmpty()) {
- if (allResults == null || allResults.size() == 0) {
+ if (allResults == null || allResults.isEmpty()) {
Thanks for the cleanup.
Thanks,
Xuelei (Andrew) Fan
On 7/20/2011 7:22 AM, Alexandre Boulgakov wrote:
Hello Sean,
Have you had a chance to look at this webrev?
Thanks,
Sasha
On 7/18/2011 6:21 PM, Alexandre Boulgakov wrote:
Hello,
Here's an updated webrev:
http://cr.openjdk.java.net/~smarks/aboulgak/7064075.2/
I've reexamined the @SuppressWarnings("unchecked")
annotations, and
added comments to all of the ones I've added. I was was also
able to
remove several of them by using covariant return types on
sun.security.x509.*Extension.get(String) inherited from Object
CertAttrSet<T>.get(String). I've also updated the consumers of
sun.security.x509.*Extension.get(String) to use the more
specific
return type, removing several casts and
@SuppressWarnings("unchecked")
annotations.
Also, please take a closer look at my changes to
com.sun.security.auth.PolicyFile.getPrincipalInfo(PolicyParser.PrincipalEntry,
final CodeSource) in
src/share/classes/com/sun/security/auth/PolicyFile.java lines
1088-1094. The preceding comment and the behavior of
Subject.getPrincipals(Class<T>) seem to be more consistent
with the
updated version, but I wanted to make sure.
The classes where I added serialVersionUID's are either new or
have
the same serialVersionUID as the original implementation.
Thanks,
Sasha
On 7/18/2011 5:33 PM, Brad Wetmore wrote:
(Apologies to folks without access to the older sources.)
On 07/18/11 15:00, Sean Mullan wrote:
On 7/18/11 5:35 PM, Alexandre Boulgakov wrote:
Is there an easy way to see when a class was added to the
JDK?
For standard API classes, you can use the @since javadoc tag
which
will indicate
the release it was first introduced in.
Standard, exported API classes. Some of the underlying support
classes for API packages like java.*.* weren't always @since'd
properly.
For internal classes, there is no easy way, since most don't
have an
@since tag.
I would probably write a script that checks the rt.jar of
each of
the JREs that
are archived at /java/re/jdk. The pathnames should be fairly
consistent, just
the version number is different.
Don't know which classes you're talking about here, but some
classes
started out in other jar files and eventually wound up in
rt.jar.
Also, some files live in files other than rt.jar. I usually
go to
the source when looking for something. If it's originally from
JSSE/JGSS/JCE, you'll need to look on our restricted access
machine.
When I'm looking for something that is in the jdk/j2se
workspaces, I
go right to the old Codemgr data, specifically the nametable
file,
because many times the files you want may be in a
src/<arch>/classes
instead of src/share/classes.
% grep -i SunMSCAPI.java
<RE-repository>/5.0/latest/ws/j2se/Codemgr_wsdata/nametable
% grep -i SunMSCAPI.java
<RE-repository>/6.0/latest/ws/j2se/Codemgr_wsdata/nametable
src/windows/classes/sun/security/mscapi/SunMSCAPI.java
ada8dbe4
a217f6b0 6c833bd3 d4ef32be
That will usually give you a good starting point.
Brad
Depending on rt.jar or not.
Chris, do you have any other ideas?
--Sean
