Review request: 8003935: Simplify the needed includes for using Thread::current() (original) (raw)

Stefan Karlsson stefan.karlsson at oracle.com
Sun Nov 25 23:10:25 PST 2012

Thanks, David.

StefanK

On 2012-11-26 08:07, David Holmes wrote:

Thanks for clarifying that Stefan. Looks okay to me.

David On 26/11/2012 5:05 PM, Stefan Karlsson wrote: On 2012-11-23 13:22, David Holmes wrote:

Hi Stefan,

I never did like these. I have to wonder though, why isn't it "runtime/thread.hpp" that all the client code #includes? Typically, you place inline functions in a inline.hpp file instead of the hpp. This helps reducing the number of unnecessary includes from spreading through out the code base. Only the users of the inline functions will include the inline.hpp file, and bring in the needed includes to implement the function. Preferably, you should only include inline.hpp files from other cpp files or other inline.hpp files. And then why isn't it thread.hpp that includes the platform specific header? In this specific case, there's an inline function in threadsolaris.inline.hpp that Thread::current() needs. StefanK David On 23/11/2012 9:54 PM, Stefan Karlsson wrote: http://cr.openjdk.java.net/~stefank/8003935/webrev

Today, whenever we use Thread::current() we have to all these lines: #include "runtime/thread.hpp" #ifdef TARGETOSFAMILYlinux # include "threadlinux.inline.hpp" #endif #ifdef TARGETOSFAMILYsolaris # include "threadsolaris.inline.hpp" #endif #ifdef TARGETOSFAMILYwindows # include "threadwindows.inline.hpp" #endif #ifdef TARGETOSFAMILYbsd # include "threadbsd.inline.hpp" #endif This patch hides this dispatching in a new file named thread.inline.hpp. Now we only have to include thread.inline.hpp. Some background to these includes: This type of dispatching was introduced into the source files when we removed the includeDB files. We discussed whether we should use "dispatch files" or not for this kind platform dependent includes. The decision was to not create dispatch files for all types of platform at that time, but instead manually create them when we thought it was warranted. thanks, StefanK

More information about the hotspot-dev mailing list