Hi,
looks good to me, although part of me can't help wonder if we'd profit
ever so little from rearranging if statements so that we test for UTF8
first and move the should-be-asserts tests for NO_ENCODING_YET down.
/Claes
On 2018-02-28 15:45, Roger Riggs wrote:
Hi,
In an effort to untangle some of the issues with property
initialization I was looking
at the platform encoding initialization and found a simplification.
Currently, the initialization occurs as a side effect of the first
call to JNU_NewStringPlatform and
involves a upcall to get sun.jnu.encoding from the system properties.
The value is cached for later use.
The native System.initProperties determines the platform specific
encoding via java_props_md.c and does an upcall to set the
sun.jnu.encoding system property, taking care to do it before the
first string that needs platform encoding.
The change directly initializes the platform encoding fast path before
it is needed to encode platform strings.
And moves the setting sun.jnu.encoding system property after the
command line arguments are inserted
keeping it from being overridden by -D on the command line that can
only confuse confusion
with code that later reads the property.
Please review and comment.
webrev:
http://cr.openjdk.java.net/~rriggs/webrev-simplify-jnu-8198697/
Issue:
https://bugs.openjdk.java.net/browse/JDK-8198697
Thanks, Roger
