[11u] RFR 8224589: Improve startup behavior of SecurityProperties

[Ivan Gerasimov]

Hi Aleksey!

On 8/26/19 10:24 AM, Aleksey Shipilev wrote:
> Original RFE:
>    https://bugs.openjdk.java.net/browse/JDK-8224589
>    https://hg.openjdk.java.net/jdk/jdk/rev/e2c952c7ff20
>
> Unfortunately, the backport to 11u is quite messy.
>
> First of all, SecurityProperties.java was added by JDK-8207768 (Improve exception messages during
> manifest parsing of jar archives) in 12u. That change is not easily backportable, because it has
> compatibility implications.
>
> Then, SecurityProperties.java was improved by JDK-8211860 (Avoid reading security properties eagerly
> on Manifest class initialization) in 12. Since the base JDK-8207768 is not backportable, this is not
> backportable as well.
>
> Then, we have this particular change. And, it has 11.0.5-oracle backports. So, what is there? I
> would guess it applies the relevant bits to the SocketExceptions.java, the block that was moved away
> by JDK-8207768:
>    https://hg.openjdk.java.net/jdk/jdk/rev/6ffa38b8da65#l4.1
>
> I think the big idea is to provide the fast path when SecurityManager is disabled. The final version
> can be seen in jdk/jdk here:
>
> https://hg.openjdk.java.net/jdk/jdk/file/tip/src/java.base/share/classes/sun/security/util/SecurityProperties.java
>
> ...which gives us the Frankenstein of 11u backport:
>    https://cr.openjdk.java.net/~shade/8224589/webrev.11u.01/
>
> Claes, Ivan, does it capture what Oracle did?


Yes, the backport looks correct.

With kind regards,

Ivan


> Aside: we can bump this backport to 11.0.6, as it is not critical.
>
> Testing: tier1
>
-- 
With kind regards,
Ivan Gerasimov