wrapper-windows-x86-64-3.5.15.zip

Java Service Wrapper Revision History. -------------------------------------- 3.5.15 * Add a new _WRAPPER_CONF_OVERRIDE setting to the Wrapper dedicated command batch files on Windows so it is now possible to control whether or not the first parameter is the configuration file name. The ability to specify an alternate configuration file is now disabled by default as it was confusing for users who tried to pass other parameters to the JVM. * Correct a couple log messages in the WrapperManager class which were missing the correct prefix identifying where they originated. * Remove some old reflection code needed for Java 1.2.x support as we have required Java 1.4 since version 3.4.0. * Remove some code to try to reconnect the backend socket from Java. It has never been possible to do so without restarting the JVM, and the related messages were confusing. * Add a new wrapper.disable_forced_shutdown property to make it possible to disable the feature to forcibly kill the JVM on shutdown if CTRL-C was pressed twice. * Reduce the number of times thread priorities are changed within the WrapperManager class to simplify the startup and shutdown process. * Fixed a dangling pointer problem, which could cause undefined behaviour whenever a property contained an unset environment variable. * Fix a race condition in the timer thread, which could cause a sigkill being propagated through the whole process group rather than the timer thread. This can only happen during the shutdown of the Wrapper. * When a child process, which got launched by WrapperManager.exec() failed to start due to a runtime-error (such as missing privileges), the forked heap persisted and the child process never finished until shutdown/restart of the JVM. The error only appears on Unix platforms when using the FORK_EXEC start-type. * Change log level and message if a certificate check returned a problem, which is not directly caused by the signature of the Wrapper, but the signature chain. * Fix a problem when the silent query command wasn't returning the correct exit code on windows Vista (and later) when the command was run from an unelevated console. Thanks to Darren for pointing this out. * The java system property wrapper.backend.so_timeout was ignored if it was set to 0, making it not possible to explicitly set the timeout to be indefinitely. * Added the properties wrapper.jvm.additional.auto_bits.<platform> to individually turn on/off the feature for the supported platforms. * Fix a problem where the script was trying to use the 64-bit binaries on Mac OSX even if the CPU was only a 32-bit architecture. This only affected versions of Mac OSX greater 10.5.0, the vast majority of those machines are already 64-bit CPU's. * The Wrapper when reloading the configuration file, was trying to access data from the call stack of a function which was actually outside of the memory range of the stack. This access violation might yield a segmentation fault. This issue was introduced in 3.5.5. Thanks to Lincoln for helping finding this problem. 3.5.14 * Fix a problem in the AppCommand.bat.in file where a parenthesis in the file name of the Wrapper binary would have caused a "PATH was unexpected at this time" error. * (Standard, Professional) Fix a problem when using a localized version of the Wrapper on Windows 64-bit platforms where the Wrapper would continue to use the default system language even wrapper.lang was used to specify a different language. Introduced in 3.5.12. * Fix a problem in the Windows AppCommand.bat.in command based batch file where the 'status' command was incorrectly being reported as 'query' in the usage output. The 'status' command had always worked correctly if used. * Fix a problem on UNIX platforms where some asynchronous messages were causing a warning message "Coding Error..." to be logged in place of the intended message. This could be seen if the configured log file did not have write permissions. Other than the incorrect log message, the Wrapper worked correctly. Introduced in 3.5.2. * Fix a problem in the UNIX script where running with upstart wasn't working correctly when RUN_AS_USER was set. * Relax security checks when running the 'status' command against the UNIX shell script so it now allows any user running the script to perform the read-only check of the pid file. * Fix a problem with the UNIX script where the 'remove' command was trying to stop a running application even when the application had not been installed. * Fix a buffer overflow which could potentially cause a crash during the installation of a Windows Service when wrapper.ntservice.account was specified. This was introduced in 3.5.12. * Fix a heap corruption which could occur on startup and potentially cause a crash. Only Windows systems, which use the System Event logs, were affected. Discovered from a code review, there had never been any reports of this causing problems for users. This could happen if the configured wrapper.log could not be written to as the Wrapper always tries to write to the Event Log in such cases. Introduced in 3.5.12. * Add a new version comparison between the UNIX shell script and Wrapper to start showing a warning in case of a version mismatch. The check will only work if the shell script and Wrapper are each of at least version 3.5.14. * Added a new wrapper.pidfile.strict property which will tell the Wrapper not to start if the pid file already existed. Defaults to false for backwards compatibility. * Make the Java side of the backend socket more resilient in case of a read or write timeout. The backend socket does not have a timeout set by default so this should not have been an issue. A couple users reported problems on specific systems however which led to this fix. * To aid in the testing of the backend socket timeout, a new wrapper.backend.so_timeout system property was added to make it possible to configure the backend socket to use a timeout. See the Javadocs of the WrapperManager class for details. 3.5.13 * Fix a typo in the script where the environment variable 'TR_BIN' should actually be 'TREXE'. This was causing the "install" command on UNIX platforms to fail. Introduced in 3.5.12. * Fix a heap corruption which could lead to a crash that would occur the second time an internal buffer used for logging was expanded. The buffer would be expanded the first time a log line over 2048 characters in length was encountered. Then the second expansion would happen when a line at least 1024 characters longer was encountered. Introduced in 3.5.11. Bug ID #3423108 3.5.12 * Put more descriptive Text in case the Wrapper is using integration method 4, but the jar file deos not specify the Main-Class correctly in its meta information. * Fix a bug when failing to grant the LogOnAsService permission to a domain user. * Fix a bug where the ident for the syslog on Unix platforms was broken since 3.5.0. This was because when opening the syslog, the Wrapper was freeing the memory for pointing to ident. However the string pointer ident will be retained internally by the Syslog routines. And must not free the memory that ident points to. Bug #3404978. * Add a check on the script to make sure the 'tr' command exists on Unix platforms. * Improve the parsing of log formats so that format tokens are recocognized even if they are lower case. This affects the wrapper.console.format, wrapper.event.default.email.maillog.format, wrapper.logdialog.format, and wrapper.logfile.format properties. * The Wrapper parses log formats by looking for known tokens, any invalid tokens are simply ignored. If the entire format is made up of invalid tokens then this r