On Feb 10, 2011, at 11:50 PM, Igor Nekrestyanov wrote:
Change itself seems ok to me.
Thanks.
In longer term we should add some "caching" support for fullpath/etc
macros, so we do not evaluate those every time
make cd to new directory ...
Yup.
In deployment makefiles we populate file with resolved variables
once (to be used with ant, not to solve cygwin problem)
and then include it every time we need them.
FWIW, we use simple rules to update the file:
0) file is generated no more than once per make run
1) file does not exist => generate it
2) file exist but make is run from top level => wipe it and
regenerate
3) file exist but explicit force parameter is passed => wipe and
regenerate
4) otherwise use existing file ...
The motivation behind it is that config is changed rarely for
partial builds and if we need to compile 1 file than waiting 30s for
makefiles to be loaded is too much.
Would be great if we can do something like this for full build ...
I agree this would be a good idea, and probably a performance
improvement for windows builds.
Just a matter of finding the time to make it happen.
-kto
-igor
On 2/10/11 1:45 PM, Kelly O'Hair wrote:
Need reviewer
7012644: Regression: jdk/make/common/shared/Defs-windows.gmk has
problems on cygwin
http://cr.openjdk.java.net/~ohair/openjdk7/jdk7-tl-cygwin/webrev/
I'm still testing builds on all platforms to make sure this is all
ok.
I will push this to both the jdk7/build/jdk and jdk7/tl/jdk
repository once it's reviewed and I'm
sure it doesn't cause any regressions to the build for RE.
-kto
