Hi Ralf,
I'm just catching up on submissions again after a Coverity squishing
expedition at the end of last week.
I'm currently look at your commit:
https://github.com/rhabacker/osg/commit/33aca40d00b8de692bc5858a15f8a53833c7d85c
Most of the changes look fine to me, the only odd thing is that the
date for the commit on this page is "committed on 15 Jan 2014" rather
than 4th of June it looks like it was pushed to your osg github
repository.
Something odd going on with dates?
--
On the issue about osgviewer not robustly handling -O, the options is
something that has been of add on rather than a full blown
implementation. The command line parsing that osgviewer and the rest
of the example provide is not meant to be a be all and end of command
line parsing. Even osgviewer is really just an example. Given this
I'm not too worried about it not supporting all the combination you
might be able to come up with w.r.t individual O options.
Personally I'd not try to make things too sophisticated, examples are
example not full blown applications. The main issue I see with the
examples not handling complex combinations of option usage is in
testing of the new feature of pairing objects in the cache with
Options. This feature is very specific to your application so far -
perhaps down the line others will start taking advantage of this
extended capability but for now it's not something that will affect
other users.
I believe the code changes in your commit are the best way to hand the
object cache so am happy to merge, but changes to the command line
parsing I'm happy to leave as it is originally, not general purpose
but at least provides a crude means of passing in options.
Robert.
On 6 June 2016 at 11:28, Ralf Habacker <[email protected]> wrote:
> Am 03.06.2016 um 11:30 schrieb Ralf Habacker:
>> Am 02.06.2016 um 13:01 schrieb Robert Osfield:
>>> One would simply use a osg::clone(options) to generate the options
>>> object to store in the cache, this would insulate the cache from being
>>> invalidated by changes to the original Options object passed in,
>> Got it, thanks :-)
>>
>> I added initial support to class ObjectCache, see
>> https://github.com/rhabacker/osg/commit/e3539ea0185a4013514850c3577180aaa38c895d
>>
>> While trying to test the implementation with osgviewer it turns out that
>> it is not possible to setup a testcase using command line.
>>
>> I used
>>
>> osgviewer -O convertToFeed test-file1.flt -O convertToKilometers
>> test-file2.flt
>>
>> and got always empty options.
>>
>>
>> Adding test1.flt with options '' to ObjectCache 0x15fadb0
>>
>> A quick check shows that in Registry::readCommandLine() all -O options
>> are read and saved in one Options instance in Registry::_options.
>>
>> void Registry::readCommandLine(osg::ArgumentParser& arguments)
>>
>> ...
>> while(arguments.read("-O",value))
>> {
>> setOptions(new Options(value));
>> }
>>
>> So I thought there may be a change to get options using one file:
>>
>> osgviewer -O convertToFeed test-file1.flt
>>
>>
>> but still got
>>
>> adding test1.flt with options '' to ObjectCache 0x12b9db0
>>
>> Any idea what's going wrong and how to fix ?
> Beside the that the mentioned command line should be
>
> osgviewer -O convertToFeed test.flt
>
> to match to debug output
>
>
> and an incomplete patch from my side, which is fixed in
> https://github.com/rhabacker/osg/commit/33aca40d00b8de692bc5858a15f8a53833c7d85c
>
>
> it looks that there are conflicts in osgviewer command line handling relating
> to the object cache.
>
>
> On command line the object cache is enabled with --enable-object-cache, which
> is handled in osgviewer.cpp:89 ff [2]
>
> // Enable caching?
> while (arguments.read("--enable-object-cache"))
> {
> if (osgDB::Registry::instance()->getOptions()==0)
> osgDB::Registry::instance()->setOptions(new osgDB::Options());
>
> osgDB::Registry::instance()->getOptions()->setObjectCacheHint(osgDB::Options::CACHE_ALL);
> }
>
>
> Adding -O <options> to command line in frot of a filename is handled in
>
> void Registry::readCommandLine(osg::ArgumentParser& arguments)
>
> ...
> while(arguments.read("-O",value))
> {
> setOptions(new Options(value));
> }
>
> The problem here is that the newly created option overwrites CACHE_ALL
> setting from [2].
>
> Second this code only adds the latest -O option and overwrites the previous
> one, which breaks command line using multiple -O options like
>
> osgviewer -O <options1> file1 -O <option2> <file2>
>
> The appended sequence diagram shows the related calls.
>
> Do you have any advice how to fix this issue(s) ?
>
> Cheers
> Ralf
>
>
>
> _______________________________________________
> osg-submissions mailing list
> [email protected]
> http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
>
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
