Re: initializing extensions in daily from Master slooow

On 03/27/2012 10:30 PM, Cor Nouws wrote:
> Experienced with some daily build the last week(s) that initializing
> extensions on the first start is verry slow.
> Known ? Issue ?

On #libreoffice-dev:
> Mar 19 16:55:09 <CorNouws> hi - initializing extensions for the first start takes longg  (in daily 
> Linux_x86 etc) Known, or should I file an issue ?
> Mar 19 17:37:19 <sberg>    mmeeks, looks like Cor's "initializing extensions for the first start takes 
> longg [sic]" is due to that sync() in ucb/source/ucp/file/shell.cxx (which, btw, I just pushed to -3-5, too)
> Mar 19 18:09:42 <mmeeks>   sberg: oh ! :-)
> Mar 19 18:10:00 <mmeeks>   sberg: I'd be amazed if a few fsync's take longer than starting Java ;-)
> Mar 19 18:10:15 <mmeeks>   but possibly we're going sync-crazy ;-) 10ms per fsync, do we really do 
> 100 of them ?
> Mar 19 18:12:02 <mmeeks>   sberg: is there an issue number ?
> Mar 19 18:12:19 <mmeeks>   I get 93 fsync's on 1st-start, seems a bit extreme but ...
> Mar 19 18:12:24 <sberg>    mmeeks, for my --enable-ext-whatever build I saw quite a number of 
> fsyncs on startup (should be much less if you have only the few bundled extensions; can be many more if 
> you have lots of shared ones)
> Mar 19 18:12:29 <mmeeks>   I lie, 142 ;-)
> Mar 19 18:12:44 <mmeeks>   on 2nd start - none in the startup path I think.
> Mar 19 18:12:56 <mmeeks>   sberg: is there a bug # ?
> Mar 19 18:13:09 <mmeeks>   sberg: [ and thanks for the back-port - you make a lot of Ubuntu users 
> happy ;-]
> Mar 19 18:13:14 <sberg>    mmeeks, issue number: no idea, Cor asked here about an hour ago (and 
> left meanwhile)

Stephan