Re: [Jack-Devel] JackSession - feedback

On Thu, May 19, 2011 at 08:54:36AM +0100, Rui Nuno Capela wrote:
> On Thu, 19 May 2011 00:47:57 +0200, Arnold Krille
> <[hidden]> wrote:
> >On Thursday 19 May 2011 00:10:26 Rui Nuno Capela wrote:
> >>adding horror to tragedy, qjackctl wipes out everything under an
> >>existing directory after prompting for any jack-session save
> >>process. so
> >>when you agree to save and replace over that, and there's where
> >>it comes
> >>to tragedy then horror, all media content will be toast!
> >
> >You fixed this bug in qtractor, but what about other apps?
> >Wouldn't it be
> >easier to have some kind of versioning inside the qjackctls
> >session format,
> >where each version gets its own version and old versions are only
> >removed when
> >the new ones are saved okay? And apps are only responsible to
> >copy/move/link
> >all the data belonging to their session.
> >
> 
> this is kind of specific to this qtractor-session file format (.qtz
> bundle) where all live media content is unzipped below the
> jack-session subdirectory...
> 
> when you save and replace a jack-session directory you'll be wiping
> out all content either from disk (qjackctl) or from sight (pyjacksm
> renames the old directory iirc) so you'll end with a broken
> qtractor-session state file anyway.
> 
> afaict, on the jack-session protocol, applications are not
> responsible to copy, move or link their data anyway as they never
> know whether they're replacing any previously existing jack-session
> directory or not--from their pov it's all a brand new jack-session
> coming up.
> 
> a /highly/ responsible jack-session manager would do better to
> create a temporary session directory first, then call all apps to
> store their data there and only after that it would rename the old
> and new jack-session directories, or copy, move or link all stored
> application state data from thereafter. qjackctl might try this in
> the future :)

thats what pyjacksm does.
however some programs actually called jack_session_reply() before
saving their data. 
resulting in pyjacksm pulling the directory away underneath them.
it looks like i am seeing this kind of behaviour with a debian version
of yoshimi, although i dont see it in the source.
> 
> anyways, qtractor's tragic issue has been solved by regression to
> old and regular session state file format, which refers to all media
> content as found on their original file-system location, ie.
> relative to qtractor-session directory, which must *not* be the same
> as the jack-session directory at all. lesson learned :)
> 
> cheers
> -- 
> rncbc aka Rui Nuno Capela
> [hidden]
> 
> Jack-Devel mailing list
> [hidden]
> http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

-- 
torben Hohn

1305794587.24349_0.ltw:2,a <20110519084247.GB3160 at siel dot b>