Writtten() to java from native (original) (raw)

Andrew Hughes ahughes at redhat.com
Fri Aug 3 17:50:21 UTC 2012

----- Original Message ----- > Hi >> This is the change to move the implementation of > De/Inflater.getBytesRead/Written() > from native to Java. Current JDK implementation uses zlib's > zstream.totalin/out fields, > which are "uLong" type in original zlib source code. This may limit > these two counters > to 4GB on 32-bit OS [1]. These two fields are currently patched to > "long long" in JDK to > support the > 4G counting. It is suggested that it might be better to > move this > 4G > counting support to java level to remove the dependency on patching > the > original zlib > source code, which might be critical in situation that Java runtime > prefers to use OS > bundled zlib, instead of the JDK bundled zlib library. >> Here is the webrev >> http://cr.openjdk.java.net/~sherman/7188852/webrev/ >> The test case attached only tests upto 128M by default (for running > in > auto testing). > The test can be run manually with option "large" to test > 4G > de/inflating. It takes > 4 > minutes to compress/decompress 5G bytes on my linux box, guess it's > probably not > desirable to hold each/every run for so long in auto setting. >> Thanks, > Sherman >> [1] http://www.zlib.net/zlibfaq.html#faq32 >> The patch looks good to me. I've also built and tested it, both with the normal and large scenarios, and with in-tree and system zlib (the latter requiring my defines.h patch) and all seems fine.

Andrew :)

Free Java Software Engineer Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/) Fingerprint = EC5A 1F5E C0AD 1D15 8F1F 8F91 3B96 A578 248B DC07