You are viewing a plain text version of this content. The canonical link for it is here.
Posted to c-dev@xerces.apache.org by "Roger Leigh (JIRA)" <xe...@xml.apache.org> on 2019/01/21 12:54:00 UTC
[jira] [Resolved] (XERCESC-2140) Add MutexMgr for C++11 mutex
implementation
[ https://issues.apache.org/jira/browse/XERCESC-2140?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Roger Leigh resolved XERCESC-2140.
----------------------------------
Resolution: Fixed
Fix Version/s: 3.2.2
Should have been closed for 3.2.2.
> Add MutexMgr for C++11 mutex implementation
> -------------------------------------------
>
> Key: XERCESC-2140
> URL: https://issues.apache.org/jira/browse/XERCESC-2140
> Project: Xerces-C++
> Issue Type: Improvement
> Components: Utilities
> Affects Versions: 3.2.1
> Reporter: Roger Leigh
> Assignee: Roger Leigh
> Priority: Major
> Fix For: 3.2.2
>
> Attachments: 0001-StdMutexMgr-Add-C-11-mutex-manager.patch
>
>
> Xalan currently supports two mutex managers: POSIX and Windows (and NoThreads, which doesn't really count). With the advent of C++11, it's no longer necessary to use platform-specific threading facilities, since it's built directly into the standard library. The attached patch adds a StdMutexMgr which uses a C++11 mutex, and will work on Unix or Windows systems with a sufficiently new compiler. thread/mutex were implemented years ago, so all recent and not so recent systems should support it. For those that don't, it will fall back to the POSIX/Windows managers and behave like before.
>
> Options have been added to manually select the desired manager as for other options for both cmake and autoconf (standard/posix/windows/nothreads). Documented in more detail on the build page.
>
> It's tested on Linux/MacOS X/Windows with a variety of manager combinations, and all looks fine so far. Any testing/comments much appreciated. It's a compatible addition, so could go into 3.2.2 if that's acceptable, otherwise could wait for later.
>
> Looking at all of the manager implementations, one key defect (likely intentional design), is that there is zero exception safety. No currently held mutex will be released if an exception gets thrown. That could be prevented by moving to using C++11 threading entirely, and using std::lock_guard, which will automatically release locks on unwind.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org