Legacy ImageMagick Discussions Archive

TL;DR Case solved, thanks everyone who helped. It turns out to be a problem with old C compiler. At the end the problem was caused by miscompiled PopShortPixel. I tried to compile with more recent Studio compiler, but it failed on magick/deprecate.h (I guess it does not like to have 'export' with ...

Yes, I have built IM myself and because of the old environment I had to make few changes. That's why I asked where the blurring actually happens instead of blaming IM :) Anyway, at the moment I think that the compiler has some problem with compiling PopShortPixel (in magick/quantum-private.h). If I ...

Oh, and the -gaussian parameter is taken from here: http://www.imagemagick.org/Usage/advanced/#3d-logos

Yes, that will be helpfull indeed :) version 6.9.5-2 I have minimized the reproducible case to broken: convert -size 100x100 xc:black -fill white -draw 'circle 50,50 13,50' -gaussian 1x1 +matte png64:logo_mask.png https://ibin.co/2o7sDXKWcyXA.png ok: convert -size 100x100 xc:black -fill white -draw ...

Hmm, the function seems to be AcquireKernelBuiltIn in magick/morphology.c. And it seems to generate nice gaussian values. From 0.000585 at the corner to 0.001592 in the middle. They numbers are small, but they seem to fall the further one is from center. I don't have this problem latest Solaris box ...

While testing ImageMagick on Solaris 10, I found that gaussian blur is broken. Sample taken from your site, enlarged a bit: convert -size 1700x1000 xc:black -fill white -draw 'circle 500,500 130,500' -draw 'circle 1200,500 1570,500' -draw 'rectangle 500,130 1200,870' -fill black -draw 'circle 500 ...

Right, so the problem was in the libtiff library which was very old on the build environment. Sorry for the noise.

Thank you, works like a charm.

Hi, FAIL: tests/validate-formats-disk.tap 1 FAIL: tests/validate-formats-map.tap 1 FAIL: tests/validate-formats-memory.tap 1 Continuing my journey to update Solaris 10 ImageMagick, these three tests fail at anything concerning TIFF/PTIF, like test 429: PTIF/Undefined/TrueColor/8-bits... fail @ tests ...

Ok, thanks. I'm curious about what the fix will be, as floating point math is evil

Ok, I quickly pasted the patch at http://pastebin.com/PHLAj7SF. It's just to show the approach I have taken.

Hi, I am trying to compile Imagemagick 6.9.4-9 on solaris 10. There are 7 tests failing FAIL: tests/cli-colorspace.tap 12 FAIL: tests/validate-formats-disk.tap 1 FAIL: tests/validate-formats-map.tap 1 FAIL: tests/validate-formats-memory.tap 1 FAIL: tests/wandtest.tap 1 FAIL: Magick++/demo/demos.tap ...