WIP: Hurd porting - #2
Conversation
Author: Jeremie Koenig Also-by: Damien Raude-Morvan (Heavily rebased)
The current freetypecheck compares the version string: a.b.c using strcmp which fails for version numbers. Let's just skip the check for now.
Since we can obtain a javac 1.6 jar that runs on 1.5, we should be able to bootstrap jdk7 using this developer bootstrap toolchain compiled from jdk6. Setting this value lower allows us to compile the 1.7 langtools targeted at 1.5 jvm so that we can run enough of the tools using the existing 1.5 jvm to bootstrap jdk7.
This enables bootstrapping of 1.6 -> 1.7 from a 1.5 gcj that has compiled the rt.jar for 1.6
Not 100% sure this is correct but fixes the compile error.
There are duplicate symbols with jvmtiEnter.o and jvmtiEnterTrace.o
|
I've just skimmed through so far but a few things stand out:
|
|
Thanks for reviewing this. There are a myriad of different changes here, yes. I have changed my approach to using icedtea7 repo as the jdk source and only minimal changes are required to icedtea repo to make it compile. The portability changes were for GNU/kFreeBSD (basically a gnu operating system with BSD kernel), and GNU/Hurd. I copied these changes from a previous attempt by Jeremie Koenig, and extended them to complete the port. I may need to refactor everything so the Hurd port is in separate files, but hiding the Hurd changes in the Linux codepaths behind |
This is not ready yet, but I wanted to send it to you so we can discuss the changes.