The following issues have been reported with this version of PDL: - A change in perl-5.14 on how the 'x' operator works affects PDL's overload of that operator for matrix multiplication. An example that shows the problem is code like: f( ($pdl0 * $pdl1) x $pdl2 ) which gives an error. The fix is to force the element-wise operation in parentheses to be treated in scalar context rather than list context. E.g., f( scalar($p0 * $p1) x $p2 ); Please see http://mailman.jach.hawaii.edu/pipermail//pdl-porters/2012-February/004426.html for a more detailed discussion and additional alternative work-arounds. - The OO interface to the GD library fails to free memory. The problem has been fixed in the pdl git and will be available in the first post-2.4.10 CPAN developers release to come. - On MS Windows, perl 5.14 (and perhaps later), it is reported that PDL::Graphics::PLplot cannot be sucessfully built as part of the PDL build. It compiles ok, but the resultant binary files are unusable. Workaround is to build PDL first, then build PDL::Graphics::PLplot separately. - The current Ctrl-C logic in the PDL shells (pdl2 and perldl) don't work the same with the perl's new "safe signals". A workaround to the problem is to set the PERL_SIGNALS environment variable to "unsafe". See sf.net feature request #3308168 for details and any future status. - The PDL build process checks for white space in the build directory as that is known to cause problems for some modules and asks you if you wish to continue [default is yes]. NOTE: You can avoid the interactive prompts by setting the environment variable PERL_MM_USE_DEFAULT to true (e.g., 1). - The Perl debugger for perls 5.10.1 through 5.14.x has a new "feature" leading to false fails for lvalue subroutines when run under the debugger. If you need to debug such code with an affected perl version, the work around is to use an intermediate temporary variable assignment as in: $piddle->slice('0:10') .= 100; # original slice code ($tmp = $piddle->slice('0:10')) .= 100; # work around perl -d "feature" The problem is understood and it appears that the error has been replaced by a warning for perl-5.15.1. NOTE: the work- around is 100% effective---but ugly. - Changes in perl-5.16.x break some existing PDL usage of lvalue subs for this release. It is planned to incorporate the needed fixes in the next PDL release. See the perldl mailing list for status updates and other announcements. - Multiline q'' constructs in the pdl2 shell now work correctly but you'll need to avoid using the q abbreviation for quit in pdl2 (at least until the multi-line plugin in Devel::REPL can be updated appropriately). If you do get "stuck" in a q quit attempt, just type Ctrl-D to exit the shell. - The demo 3d and 3d2 windows do not close (can not be closed) after the demo finishes. You need to exit the perldl shell to have the window close. - When you close a TriD graphics window with the frame widget the whole process exits including the perldl shell. - PDL::FFTW is based on version 2 of the FFTW API. The current release of the FFTW library is version 3.2.2. To use PDL::FFTW you will need to install FFTW version 2.1.5, use the built-in PDL::FFT routines instead, or (better yet) submit a patch to update PDL::FFTW to the version 3 API. - Extremely verbose but not particularly helpful text output from the configure-build-test process. - Directory completion in the interactive PDL shells (perldl and pdl2) using Term::ReadLine::Perl adds a space after each directory expansion. To continue to complete, one must delete the space before typing again. The problem has been reported as a Term::ReadLine::Perl bug. - PDL on Cygwin has (by default) a 300MB process memory limit which can cause out of memory crashes with large data sets. See the cygwin users guide: http://www.cygwin.com/cygwin-ug-net/setup-maxmem.html for how to increase the memory limit. - minimum/maximum/min/max inconsistently handle NaN values. NaNs for the first element of a pdl will propagate to the result while if the first element is a number, then the result will be a number. The answer for max/min should not depend on the order of the input values. - 16-bit image IO via rpic/wpic and rim/wim do not correctly handle byte order issues. The intermediate PNM data piped through the converter programs is not adjusted to network byte order per the NetPBM convention/standard. This is not an issue for big-endian systems or for 8-bit/channel image data. - The following SourceForge bugs are outstanding at time of the PDL-2.4.10 release: 3479009 Slice does not like whitespace 3476648 PDL build of HTML docs fails on recent bleed Perl 3475075 t/pic_16bit.t subtest fails 3465663 perldl.conf OPTIMIZE option doesn't work 3462924 bogus "dimension mismatch" for Nx0 index piddles 3460886 t/flexraw_fortran.t test failures 3363406 PDL::Minuit build can't make libminuit.a 3316394 'help funname' fails to show multiple function names 3314708 PDL install does not install scantree.pl 3299615 PLplot still unusable with X 3234141 max() fails on nan 3161459 online docs don't link to functions in other PDL modules 3018731 NiceSlice affects code where it should not 2995500 perl -d chokes on lvalue functions 2153898 default PDL build output too long 1994584 PDL Has no signed 8-bit integer data type 1507054 closing TriD window kills perldl shell 1205359 PGPLOT Window does not store full state info For more information on these and other PDL issues, and for submissions of patches (bug patches are always welcome!), see the PDL mailing lists. Links to archive list discussions and how to register for the mailing lists can be found at http://pdl.sourceforge.net/maillists/ .