Hi Robert, Serge,
I have no problem whith dropping the submission, I agree that the
plugins should behave similar. I didn't even check all memory
allocations in the plugin, just the big block.
I'am still confused about the code just below new, (in the current svn
version) and almost sure it cannot work.
buffer = new unsigned char [w*h*format];
for(unsigned char* ptr=buffer;ptr<buffer+w*h*format;++ptr) *ptr = 0;
if (!buffer)
{ //..cleanup and return NULL
It's the last line that confused me, if new did not throw an exeption,
than the use of a null pointer in the next line will. I think (!buffer)
is bound to be false.
Laurens.
Robert Osfield wrote:
Hi Laurens and Serge,
I'm torn on this one. In general I've left it up to client code to
catch exceptions and handle them how they want. In the case of
plugins catching the exceptions and passing back an OUT_OF_MEMORY
error might be cleanest route. It wouldn't make sense to have some
plugins use this, and others not though, so if we are to make this
change, then it should be done for all plugins. There are 50+ plugins
though, so making this change is something that takes several hours of
careful and tedious work. We are very close to a stable release so
I'd suggest making such extensive changes after 2.6 goes out.
Robert.
On Mon, Jul 21, 2008 at 3:28 PM, Serge Lages <[EMAIL PROTECTED]> wrote:
Hi Laurens,
I am currently using the Tiff plugin (and others) with large files, and I
use the std::exception to catch the allocations problems (it allows me for
some process to know that there is an allocation problem, clear the object
cache, and try to load again for example). For me it's a method which works
with all plugins, but with your change, the tiff plugin will have a
different behavior, and I will not be able to know if the
ERROR_IN_READING_FILE comes from an allocation problem or not.
So I recommend to stay like now, having bad_allocations exceptions when we
cannot allocate a buffer, or to move all the plugins to check for
allocations problems and return a specific error for that. But having some
plugins returning an error and others throwing an exception can be really
confusing.
Cheers,
On Mon, Jul 21, 2008 at 4:10 PM, Laurens Voerman <[EMAIL PROTECTED]> wrote:
Now with attachment I hope...
Laurens Voerman wrote:
Hi Robert,
I made a few changes in src\osgPlugins\tiff\ReaderWriterTIFF.cpp to avoid
an unexpected program termination when using very lage textures. I an not
totally sure about the portability of "new(std::nothrow)" , I need it on my
WinXP/Visual studio 2008 Express system where the new operator throws an
exeption, instead of returning a null pointer.
Second change: return ERROR_IN_READING_FILE on failure seems more
appropriate to me.
Greetings, Laurens.
_______________________________________________
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
--
Serge Lages
http://www.tharsis-software.com
_______________________________________________
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
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
