framework, unotest: do not override UserInstallation from environment

The PythonTest_dbaccess_python may fail in incremental builds because of
stale JVM path in workdir/unittest/user/config/javasettings_Linux_X86_64.xml
so it's better to use the fresh UserInstallation that is already passed
in from PythonTest.mk.  To make that work, tolerate non-existing path in
SubstitutePathVariables::SetPredefinedPathVariables() at least when
running in the build environment, which is probably better than
effectively running with "/" as UserInstallation anyway...

This partially reverts commit 6dbb6275ebd1a4299099c3b6bc82ec4ee0e1fb86.

Change-Id: I4d42d53c049056590662828360c4ce5a29af0f4b

1 files changed, 3 insertions, 1 deletions

diff --git a/framework/source/services/substitutepathvars.cxx b/framework/source/services/substitutepathvars.cxx
index 67fde1a9e92e..3a55a2dd324a 100644
--- a/framework/source/services/substitutepathvars.cxx
+++ b/framework/source/services/substitutepathvars.cxx
@@ -1250,7 +1250,9 @@ void SubstitutePathVariables::SetPredefinedPathVariables()
     //TODO: Is that still the case? (With OOo 3.4, "unopkg sync" was run as part
     // of the setup. Then no user installation was required.)
     //Therefore we do not assert here.
-    if( aState == ::utl::Bootstrap::PATH_EXISTS ) {
+    // It's not possible to detect when an empty value would actually be used.
+    // (note: getenv is a hack to detect if we're running in a unit test)
+    if (aState == ::utl::Bootstrap::PATH_EXISTS || getenv("SRC_ROOT")) {
         m_aPreDefVars.m_FixedVar[ PREDEFVAR_USERPATH ] = ConvertOSLtoUCBURL( sVal );
     }