building part of jdk 8 (original) (raw)

Pete Brunet peter.brunet at oracle.com
Mon Feb 20 20:01:47 UTC 2012

On 2/20/12 1:25 PM, Fredrik Öhrström wrote:

Den måndagen den 20:e februari 2012 skrev Pete

> > A fastdebug build is a combination of the variables VARIANT=DBG and FASTDEBUG=true > but it's been a long time since I looked at that. Proof is in the pudding. Good Luck.

Thanks Kelly, That's true. Also, I hadn't realized that I need to use target fastdebugbuild at the top level and fastdebug at the lower levels. The new build system tries to avoid this confusion by simplifying the setup. ./configure --with-debug-level=slowdebug where debug level is release,fastdebug,slowdebug. This setting affects all repos, both hotspot and the jdk. Hi Fredrik, I can't find any files named configure. Is there further documentation? -Pete Previously the jdk build took so long time that it made sense to squeeze every version of hotspot binaries into the same hotspot build directory. (several different debug versions, even 32 and 64 bit on sparc platforms?) Since whatever you did, it was vital that you did not have to recompile the jdk. The result was the confusing pudding of files in the build directory. With the new build system, rebuilding the jdk is faster than rebuilding hotspot. You can therefore afford to create many many different configurations and build them all from the same source. This is why the configuration warns when you build more than one jvm. Typically on 32 bit platforms you want to supply --with-jvm-variants=server to only build the best jvm. By default, mimicing the old makefiles, the default on 32 bit is --with-jvm-variants=client,server,minimal resulting in a build that spends 75% of the build time building jvm variants..... //Fredrik -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://mail.openjdk.org/pipermail/build-dev/attachments/20120220/bf7b353c/attachment.htm>

More information about the build-dev mailing list