New version has been released. The ebuild includes updated SRC_URI and optional media-gfx/graphicsmagick[imagemagick] dependency as suggested by Peter Volkov: bug #314231
Have you tested it? I've segfaults on recognition attempt.
Created attachment 228993 [details] app-text/cuneiform-0.9 Updated SRC_URI, added optional graphicsmagic dependency.
Looks like the ebuild didn't upload the previous time, and I din't notice :( Works fine with ImageMagic. Havent't tested with media-gfx/graphicsmagick. Do you want me to? Also, I've noticed that 0.8.0 segfaulted on large images (over 17 Mpix). Supposedly this has been fixed by 0.9.0, but I haven't tested that either.
Have you checked it on amd64?
Yes. But not on x86. My emerge --info: http://nopaste.info/41459e2242.html
For me 0.9.0 works only without imagemagick (checked with 6.5.8.8 and 6.6.1.2).
Yaleks, could you share with us some test case to reproduce problem on amd64? (In reply to comment #3) > Works fine with ImageMagic. Havent't tested with media-gfx/graphicsmagick. Do > you want me to? Please, do.
Use cuneiform-linux-0.9.0/images/ from source tarball.
$ cuneiform -l rus clipb.bmp Cuneiform for Linux 0.9.0 $ less cuneiform-out.txt
(In reply to comment #7) > Please, do. I've got /usr/include/GraphicsMagick/Magick++.h and /usr/lib64/libGraphicsMagick*, but cmake-2.6.4-r3 does not detect them as ImageMagick. How do I need to patch cuneiform_src/CMakeLists.txt ? What else do I need to patch?
Looks like CMake searches for Magic++.h in /usr/include/ and /usr/include/ImageMagic/, while GraphicsMagic installs it to /usr/include/GraphicsMagic/. Libraries have different paths and names too. CMake has no module for GraphicsMagic. Does GraphicsMagic has options to duplicate directory structure and library names of ImageMagic? Otherwise there should be similar problem with every package using ImageMagic and CMake.
(In reply to comment #10) > (In reply to comment #7) > How do I need to patch cuneiform_src/CMakeLists.txt ? What else do I need to > patch? > Please discuss this in bug #314231. This is not a very important issue (see meta bug #314431 ).
Created attachment 231381 [details] app-text/cuneiform-0.9 ebuild Removed GraphicsMagick dependency.
version 1.0 is out.
(In reply to comment #14) > version 1.0 is out. > I know, but the same problems with ImageMagick.
yaleks, two possible solutions here: 1. Bump it hardmasked 2. Drop amd64 keyword as if I understand correctly this is amd64 related problem. What do you think? Also do you mean this bug: https://bugs.launchpad.net/cuneiform-linux/+bug/261364 ?
(In reply to comment #16) > yaleks, two possible solutions here: > 1. Bump it hardmasked > 2. Drop amd64 keyword as if I understand correctly this is amd64 related > problem. We can bump it, but mask "imagemagick" for amd64.
cuneiform 1.0.0 works for me on ~amd64 with ImageMagick 6.6.2.5 (and graphicsmagick 1.3.12) and the test files from $src/images folder and own scanned png-snippets in two languages. I did not have any compiling or segfault problems. I was able to create a buffer overflow though, with one own large testimage containing two languages and specifying the non-english one. But I guess this is another bug and will try to report it to the developers.
yaleks, could you prepare and attach here diff for something to commit then? :)
(In reply to comment #18) > I was able to create a buffer overflow though, with one own large testimage > containing two languages and specifying the non-english one. But I guess this > is another bug and will try to report it to the developers. as I found out, this is a compiler optimization problem: using -O1 or higher produces this overflow, so maybe you want to enforce -O0 in the ebuild to prevent this problem until the developers fixed it.
> $ cuneiform -l rus russian2bl.bmp Cuneiform for Linux 1.0.0 PUMA_XFinalrecognition failed. Not a segfault for me (the same with no "-l rus").
(In reply to comment #19) > yaleks, could you prepare and attach here diff for something to commit then? :) > just copy cuneiform-0.8.0.ebuild to cuneiform-1.0.0.ebuild and add ">app-text/cuneiform-0.8.0 imagemagick" to /usr/portage/profiles/arch/amd64/package.use.mask
Ok, finally new version is in the tree. Crash mysteriously went away. We suspect this was imagemagick bug, but who knows...