java-rmi.exe/cgi

[Kelly O'Hair]

See http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6512052

My gut feeling is that we fix it, and then add this file to an exclusion list when we get to the point of
comparing built jdk images.
But I'm willing to accept other suggestions.

-kto

On Nov 10, 2011, at 10:14 AM, Erik Joelsson wrote:

> Sometimes when converting makefiles in the jdk repo, I stumble over things that don't look right. The code seem to suggest one intention, but careful inspection reveals a different behavior. Normally I try to copy the actual behavior and put a note in the new make file about it. Now I discovered something that's just too weird:
> 
> In "jdk/make/sun/rmi/cgi/Makefile" the program "java-rmi.exe" is built. This looks like it's intended to be a windows native version of the "java-rmi.cgi" script that is available on linux and solaris. The makefile even copies java-rmi.exe to java-rmi.cgi. When creating the j2sdk/jre images, however, only the exe version is included. I verified this in the promoted builds of jdk 7 (b148 and a few others). The makefile also tries to declare that the program be compiled from "jdk/src/windows/bin/java-rmi.c" (by declaring FILES_c variable) but Program.gmk ignores this and just builds main.c (the standard java launcher) instead. So the "java-rmi.exe" in a windows distribution is just java.exe with a weird name.
> 
> What is this script/program used for, is it used at all, and how should I convert the building of it? I have tried compiling java-rmi.c and it seems to be working at least. (Gives same default output as the script on linux)
> 
> /Erik