diff options
Diffstat (limited to 'jvmfwk/source/elements.hxx')
| -rw-r--r-- | jvmfwk/source/elements.hxx | 28 |
1 files changed, 28 insertions, 0 deletions
diff --git a/jvmfwk/source/elements.hxx b/jvmfwk/source/elements.hxx index 93aec47718e1..6efedbf96647 100644 --- a/jvmfwk/source/elements.hxx +++ b/jvmfwk/source/elements.hxx @@ -196,6 +196,34 @@ private: */ boost::optional< ::std::vector< ::rtl::OUString> > m_JRELocations; + /** Only in INSTALL mode. Then NodeJava.write writes a <modified> element + which contains the seconds value of the TimeValue (osl/time.h), obtained + with osl_getSystemTime. + It returns 0 if the value cannot be obtained. + This is used to fix the problem that the modified time of the settings + file is incorrect because it resides on an NFS volume where the NFS + server and NFS client do not have the same system time. For example if + the server time is ahead of the client time then checkSettingsFileStatus + deleted the settings. So even if javaldx determined a Java + (jfw_findAndSelectJRE) then jfw_startVM returned a JFW_E_NO_SELECT. Then + it looked again for a java by calling jfw_findAndSelectJRE, which + returned a JFW_E_NONE. But the following jfw_startVM returned again + JFW_E_NO_SELECT. So it looped. (see issue i114509) + + NFS server and NFS client should have the same time. It is common + practise to enforce this in networks. We actually should not work + around a malconfigured network. We must however, make sure that we do + not loop. Maybe a better approach is, that: + - assume that mtime and system time are reliable + - checkSettingsFile uses system time and mtime of the settings file, + instset of using getModifiedTime. + - allow a small error margin + - jfw_startVM must return a JFW_E_EXPIRED_SETTINGS + - XJavaVM::startVM should prevent the loop by processing the new return+ value + + */ + sal_uInt32 getModifiedTime() const; + public: NodeJava(Layer theLayer = USER_OR_INSTALL); |