Nonetheless, the images look ok which suggests that the *test checks* are failing and not necessarily the PNM and PNG image input and output. Ingo will need to debug that on his end unless you can reproduce the problem on your end.
--Chris On Tue, Jan 17, 2012 at 2:53 PM, David Mertens <[email protected]> wrote: > On Tue, Jan 17, 2012 at 12:49 PM, Chris Marshall <[email protected]> > wrote: >> >> My guess is that the test is checking for equality >> rather than approximately. >> >> Could you run the test in the debugger and see >> what the value of sum(abs($a16-$a16_png)) is >> for the failing test? >> >> What are the data types of the piddles being >> compared? > > > The types are PDL ushorts, so they really should not suffer rounding errors. > Also, my hunch was wrong: shorts appear to be consistently 16 bits, as > discussed here: http://en.cppreference.com/w/cpp/language/types > > David > > >> >> Thanks, >> Chris >> >> On Tue, Jan 17, 2012 at 1:29 PM, Ingo Schmid <[email protected]> wrote: >> > >> > I still get an error, although the images look the same (s. >> > attachments), >> > according to diff. >> > >> > riococo:[tmp]$ perl pic_16bit.t >> > 1..5 >> > ok 1 - use PDL::IO::Pic; >> > Using the command NONE with the flags -quiet >> > scaling data to type byte... >> > using intermediate format PGM >> > built the command >tbyte_a.pnm to write image >> > conversion by 'tbyte_a.pnm' >> > ok 2 - pnm byte image save+restore >> > Using the command pnmtopng with the flags -quiet >> > scaling data to type byte... >> > using intermediate format PGM >> > built the command |pnmtopng -quiet >tbyte_b.png to write image >> > conversion by 'pngtopnm -quiet tbyte_b.png |' >> > ok 3 - png byte image save+restore >> > Using the command NONE with the flags -quiet >> > using intermediate format PGM >> > built the command >tushort_a16.pnm to write image >> > conversion by 'tushort_a16.pnm' >> > ok 4 - pnm ushort image save+restore >> > Using the command pnmtopng with the flags -quiet >> > using intermediate format PGM >> > built the command |pnmtopng -quiet >tushort_b16.png to write image >> > conversion by 'pngtopnm -quiet tushort_b16.png |' >> > not ok 5 - png ushort image save+restore >> > # Failed test 'png ushort image save+restore' >> > # at pic_16bit.t line 66. >> > # Looks like you failed 1 test of 5. >> > >> > Best >> > Ingo >> > >> > >> > >> > On 01/17/2012 06:57 PM, Chris Marshall wrote: >> >> >> >> Please run this modified pic_16bit.t and send the >> >> generated files. I've attached a tar file with the new >> >> test as well as the images output on my platform >> >> (which look ok). >> >> >> >> Thanks, >> >> Chris >> >> >> >> On Tue, Jan 17, 2012 at 9:25 AM, David >> >> Mertens<[email protected]> >> >> wrote: >> >>> >> >>> Ingo - >> >>> >> >>> Do I understand correctly that you *updated* the PDL ebuild? If so, >> >>> you >> >>> have >> >>> my thanks, as well as the thanks from the rest of the porters! We >> >>> should >> >>> add >> >>> instructions to our wiki for installing PDL on Gentoo. Anything that >> >>> makes >> >>> PDL easier to install on any platform should be mentioned. >> >>> >> >>> Now, as to your bug, my test passes and looks like the tushort_a16.png >> >>> image >> >>> that you attached in your follow-up. I have not worked with (or on) >> >>> PDL::IO::Pic and I do not have experience with the C library that >> >>> backs >> >>> it >> >>> up, so I cannot be of much help, at least, not with the details that >> >>> you >> >>> have posted so far. However, this looks like a bona-fide bug, so we >> >>> should >> >>> probably file a bug report. If you have a sourceforge account, I would >> >>> appreciate if you could file a bug here: >> >>> >> >>> >> >>> http://sourceforge.net/tracker/?func=browse&group_id=612&atid=100612&status=1. >> >>> If you do not have a sourceforge account, please reply with the output >> >>> of >> >>> "perldl -V" so that I can include it in the bug report. >> >>> >> >>> Last point: from the build log, it looks like you have a 64 bit AMD >> >>> chip. >> >>> I >> >>> wonder if this is an issue with 64 bit builds, especially the 64 bit >> >>> representation of PDL's short type. But that's just a hunch. >> >>> >> >>> David >> >>> >> >>> -- >> >>> Sent via my carrier pigeon. >> >>> >> >>> _______________________________________________ >> >>> Perldl mailing list >> >>> [email protected] >> >>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl >> >>> >> > >> >> _______________________________________________ >> Perldl mailing list >> [email protected] >> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl > > > > > -- > Sent via my carrier pigeon. _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
