2008.10.14 18:00 "[Tiff] debugging on Windows", by Rajmohan Banavi

2008.11.17 17:20 "Re: [Tiff] Memory leak (TIFFOpen, TIFFReadTile)?", by Frank Warmerdam

I have found that memory-mapped I/O is completely FUBARed, even on Linux. Even if the kernel is ultimately going to release the memory, the system still thrashes like crazy. It really shouldn't be the default mode for TIFFOpen().

Folks,

I am interested in advice on whether to change TIFFOpen() to *not* use memory mapped IO by default. Presumably the application would need to pass "M" in the open options if they wanted memory mapping when possible.

If there is consensus that the change would be good we can apply it in 4.0 and 3.9 branches.

I personally mostly use libtiff through my own IO layer which does not offer memory mapping so I have little recent experience with the benefits or problems of memory mapped io.

Best regards,
--
---------------------------------------+--------------------------------------

I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam@pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam

and watch the world go round - Rush    | Geospatial Programmer for Rent