Sorry, I've only just noticed this message sitting in my to-do pile. An interesting point because , yes, checkouts are not atomic. It wouldn't be hard to extend the code to retrofit atomicity around the svn code using directory renames and the like, but before we put the effort into the code I've got one question: Assuming we consistently always do atomic incremental updates to individual directories, do we store the data in the directories such that an arbitrary combination of directory versions is likely to be safe and legal? I believe not.
If not, terrasync-atomic has to start by replacing your flightgear-data with a completely empty directory and only, as individual paths are verified against SVN, rename the existing (maybe updated) content into it. This would seem to be exciting, both in terms of how long the directory stays empty during initial verification (so the simulator is unusable) and in terms of how to decide what the initial prerequisites are (so the simulator doesn't get confused). I'm tempted to suggest that we either have to commit to a dependency graph for the directories that terrasync can use (and flightgear can verify at runtime), or flightgear's data loading routines need to become terrasync aware (so their thread can block while waiting for the sync to complete). If someone has a cleaner idea for how to infer the dependency graph among our many different types of directory, I'd love to hear it. John Denker wrote: > Hi -- > > When terrasync is running, are the updates atomic? > I suspect not, since terrasync depends on svn, and > AFAIK svn commits are atomic but checkouts are not. > > I've seem some pretty weird irreproducible results > which might be explained by FG reading half of a > new file plus half of an old file because terrasync > was in the middle of an update. > > So far this is mostly just a hypothesis, but it fits > the facts, and I haven't been able to come up with > any other hypotheses to explain what I've seen. > > As an example, FG died via an assertion in navradio. > The runway was numbered "0". Needless to say, the > airport in question doesn't have any such runway. > > === > > I hear rumors of a major upgrade/rewrite to terrasync. > This might be something to keep in mind. > > ------------------------------------------------------------------------------ > The Planet: dedicated and managed hosting, cloud storage, colocation > Stay online with enterprise data centers and the best network in the business > Choose flexible plans and management services without long-term contracts > Personal 24x7 support from experience hosting pros just a phone call away. > http://p.sf.net/sfu/theplanet-com > _______________________________________________ > Flightgear-devel mailing list > Flightgear-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/flightgear-devel > ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
