Erik:
> The fix is to make sure WORKSPACE_ROOT in basics.m4 only gets values
> that have been fixed using UTIL_FIXUP_PATH.
Looks good.
Tim
This is a fix for the problem reported by Ty in this thread:
https://mail.openjdk.java.net/pipermail/build-dev/2020-July/027917.html
The problem is the --disable-absolute-paths-in-output option, which is
default set to disable on release builds. When not allowing absolute
paths, we rewrite absolute paths to relative in several types of build
command lines. This rewrite relies on the WORKSPACE_ROOT variable to be
fully resolved. Currently, if the workspace is located inside the cygwin
root dir, the WORKSPACE_ROOT variable will have a path looking like
"/home/user/jdk" while all other paths that we resolve files from will
look like "/cygdrive/c/cygwin64/home/user/jdk". This causes the rewrite
to fail and results in build errors like this:
c1xx: fatal error C1083: Cannot open source file:
'../../../..c:/cygwin64/home/young/jdk-master/jdk-master/build/windows-x86_64-server-release/hotspot/variant-server/libjvm/objs/BUILD_LIBJVM_pch.cpp':
No such file or directory
The fix is to make sure WORKSPACE_ROOT in basics.m4 only gets values
that have been fixed using UTIL_FIXUP_PATH.
(This issue rarely occurs within Oracle, where Jib is already changing
the working to be on the /cygdrive format, which is why we haven't seen
it reported internally so far.)
Bug: https://bugs.openjdk.java.net/browse/JDK-8249255
Webrev: http://cr.openjdk.java.net/~erikj/8249255/webrev.01/index.html
/Erik
