I got similar problem on Windows 2003 server. After adding printf to hyenv.c, found that the following weird vars are reported by GetEnvironmentStrings() Win32 API:
... [junit] Next var==::=::\ [junit] Next var==C:=C:\.usr\varlax\harmony\trunk\working_vm\build [junit] Next var==ExitCode=00000000 [junit] Next var==G:=G:\avarlamo\harmony [junit] Next var=ALLUSERSPROFILE=C:\Documents and Settings\All Users [junit] Next var=ANT_CMD_LINE_ARGS=-f make/build.xml -Djava.debug.option on -DBUILD_CFG debug kernel.test -Dtest.case java.lang.SystemTest [junit] Next var=ANT_COMMAND=C:\apache-ant-1.6.5\bin\ant.bat [junit] Next var=ANT_HOME=C:\apache-ant-1.6.5 ... XSI [1] prohibits '=' signs in key names, but clearly MS does not care... So - what to do? I wonder why this did not happen when DRLVM used apr_env_get... [1] http://www.opengroup.org/onlinepubs/007908799/xbd/envvar.html 2006/12/4, Vladimir Ivanov <[EMAIL PROTECTED]>:
The test java.lang.SystemTest failed for my on windows XP. My investigation shows that for environment variable: !C:=C:\harmony.top\drlvm\trunk\build\win_ia32_msvc_debug\semis\kernel.tests\classes Values for RI and Harmony are differing: RI: key = =C: env.get(key)= C:\harmony.top\drlvm\trunk\build\win_ia32_msvc_debug\semis\kernel.tests\classes System.getenv(key) = C:\harmony.top\drlvm\trunk\build\win_ia32_msvc_debug\semis\kernel.tests\classes Harmony: key = env.get(key)= C:=C:\harmony.top\drlvm\trunk\build\win_ia32_msvc_debug\semis\kernel.tests\classes System.getenv(key) = null Actually, I'm not sure that it is originally "!" sign is met in the variable while I can't reproduce it through the set !C:=C:\junit.jar Could somebody else reproduce/ investigate this failure too? Thanks, Vladimir
