| 060d207 | Pascal Obry | 03 March 2016, 17:14:54 UTC | dt_dev_distort_[back]transform_plus add exception for module. In these routines we had a way to select the range of module priority to use to apply the distortion. A new parameter except is introduced to be able to add an exception for a module to skip during the computation. If the parameter is NULL there is no exception and the routines are behaving as before. This is needed and used in liquify as the distortion for the control elements must be computed for the whole pipe (before liquify we have lens and ashift to take into account, and after roration and crop) except liquify itself which must not modify the control point. Proper fix for #10939. | 03 March 2016, 17:14:54 UTC |
| 26a1ad6 | Pascal Obry | 03 March 2016, 09:13:04 UTC | Revert "liquify: gui_post_expose should distort points through all iop." This reverts commit 038ed96b20232bfe35035cc6d74d156431b48ab9. The proper fix would be to distort the controls through all modules except liquify. We do not want the control to be distorted, the control must stays at the position it has been placed by the user. Note that the gui_post_expose and the get_point_scale routines must distort the points through the same modules. For #10939 | 03 March 2016, 09:35:13 UTC |
| 590ceed | Roman Lebedev | 03 March 2016, 08:53:18 UTC | Merge pull request #1162 from LebedevRI/cmake-issues CMake fixes: part 2. Fixes #10354 | 03 March 2016, 08:53:18 UTC |
| db08ec3 | Pascal Obry | 02 March 2016, 22:06:28 UTC | Update French translation. | 02 March 2016, 22:06:28 UTC |
| e386cce | Tobias Ellinghaus | 02 March 2016, 19:57:14 UTC | Sort parsed GPX points by time Later when using the data we assume that they are sorted. | 02 March 2016, 19:57:14 UTC |
| c1ba149 | AlicVB | 02 March 2016, 18:39:32 UTC | Merge pull request #1153 from AlicVB/coll_pathsort collections: add sort by full path | 02 March 2016, 18:39:32 UTC |
| da22494 | Roman Lebedev | 02 March 2016, 15:18:17 UTC | dt_colorspaces_get_output_profile(): also check for get_p() | 02 March 2016, 15:18:17 UTC |
| 9a03679 | Roman Lebedev | 02 March 2016, 14:49:35 UTC | dt_colorspaces_get_output_profile(): do not crash when there is no colorout | 02 March 2016, 14:49:35 UTC |
| 9c28b56 | Roman Lebedev | 02 March 2016, 14:28:18 UTC | dt_view_image_expose(): do not leak rgbbuf when surface == NULL | 02 March 2016, 14:28:18 UTC |
| bc9b599 | Roman Lebedev | 02 March 2016, 14:13:49 UTC | Modulelist lib: get_params(): when realloc() fails, orig ptr is intact. do not leak it. | 02 March 2016, 14:13:49 UTC |
| 179a750 | Roman Lebedev | 02 March 2016, 13:58:56 UTC | Format PDF: fix scan-build "Allocator sizeof operand mismatch" warning. | 02 March 2016, 13:58:56 UTC |
| d8de087 | Roman Lebedev | 02 March 2016, 13:51:40 UTC | free() and g_free() can handle NULL. | 02 March 2016, 13:51:40 UTC |
| 2552af4 | Tobias Ellinghaus | 02 March 2016, 11:53:30 UTC | [lua] Expose index of selected combobox entry | 02 March 2016, 11:53:30 UTC |
| e992daa | Tobias Ellinghaus | 02 March 2016, 11:30:55 UTC | [lua] support slider's step and digit settings | 02 March 2016, 11:30:55 UTC |
| ccd3120 | Tobias Ellinghaus | 02 March 2016, 10:22:48 UTC | Build EN manual/lua-api in the catch-all rule, too ... and not just the localized versions. | 02 March 2016, 10:22:48 UTC |
| a3c09d2 | Tobias Ellinghaus | 02 March 2016, 09:53:00 UTC | Don't use newlines in dt_control_log | 02 March 2016, 09:53:00 UTC |
| 8f42a2a | Ulrich Pegelow | 02 March 2016, 05:53:00 UTC | Merge pull request #1156 from dtorop/fcxtrans-rework FCxtrans rework | 02 March 2016, 05:53:00 UTC |
| a762de3 | Dan Torop | 02 March 2016, 05:24:37 UTC | demosaic: fix allhex declaration some versions of gcc notice a type mismatch in 3rd argument of allhex(). | 02 March 2016, 05:24:37 UTC |
| 6bdfc81 | Tobias Ellinghaus | 02 March 2016, 00:00:54 UTC | [l10n] Partial update of German translation | 02 March 2016, 00:00:54 UTC |
| 1564423 | Tobias Ellinghaus | 01 March 2016, 23:11:18 UTC | Show accel in button tooltips & some accel cleanup We already mentioned some accels in the tooltips but hardcoded them to the initial default values (and in the case of invert-selection it was the wrong one) so it didn't reflect the actual user settings. Now the right keys should be shown. I had to move some accels to other libs and hope that it didn't break anything. | 01 March 2016, 23:11:18 UTC |
| 0b74cbd | AlicVB | 01 March 2016, 20:29:05 UTC | collect: safeguard against malformed db values | 01 March 2016, 20:31:53 UTC |
| 88877a2 | Roman Lebedev | 01 March 2016, 18:43:25 UTC | Presets: fix build: gtk_widget_set_tooltip_text() takes widget, not GObject | 01 March 2016, 18:43:25 UTC |
| 026311d | Roman Lebedev | 01 March 2016, 16:12:06 UTC | Nlmeans iop: add non-SSE codepath | 01 March 2016, 18:09:32 UTC |
| f878be9 | Roman Lebedev | 01 March 2016, 13:13:52 UTC | Nlmeans iop: copy process_sse2() as process() | 01 March 2016, 18:09:32 UTC |
| 4c70582 | Tobias Ellinghaus | 01 March 2016, 17:52:45 UTC | Allow printing with ctrl-p shortcut This was already advertised in the tooltip of the button but the actual code was missing ... | 01 March 2016, 17:52:45 UTC |
| 4d77940 | Tobias Ellinghaus | 01 March 2016, 17:40:28 UTC | Small cleanup of repetitive code in colorbalance | 01 March 2016, 17:40:28 UTC |
| 90d0a08 | Tobias Ellinghaus | 01 March 2016, 17:38:07 UTC | Use gtk_widget_set_tooltip_text() This feels cleaner than using g_object_set() and it's easier to remember and less confusing. | 01 March 2016, 17:38:07 UTC |
| 301affb | Roman Lebedev | 29 February 2016, 19:29:52 UTC | Bilateral{,cl}: add proper header, #include it, not .c AKA: Iops: enforce public API: add some static keyword. Part 2. Now none of our libs (not counting libdarktable) export any symbols which are not part of it's public API. | 29 February 2016, 19:29:52 UTC |
| ef1b0fe | Roman Lebedev | 29 February 2016, 18:47:41 UTC | Rename bilateral{,cl} .h -> .c | 29 February 2016, 18:47:41 UTC |
| 639d97a | Roman Lebedev | 29 February 2016, 17:05:55 UTC | CMake: unbreak git version string update :) Unlike add_custom_target(), add_custom_command() runs only when one of it's DEPENDS is changed, so naturally, it broke version string regeneration. So will have to live additional 'proxy' level to ensure that cmake is happy (add_custom_command with OUTPUT) AND that the version string is actually updated... | 29 February 2016, 17:05:55 UTC |
| 81721d3 | Roman Lebedev | 29 February 2016, 15:40:03 UTC | Iops: enforce public API: add some static keyword. Part 1. Part 2 - need to deal with dt_bilateral_*()... | 29 February 2016, 15:40:03 UTC |
| dec6630 | Roman Lebedev | 28 February 2016, 14:40:55 UTC | Libs: enforce public API: add some static keyword. | 28 February 2016, 14:40:55 UTC |
| 33efa4a | Roman Lebedev | 28 February 2016, 14:32:18 UTC | Imageio storage: enforce public API: add some static keyword. | 28 February 2016, 14:32:18 UTC |
| 22bd45c | Roman Lebedev | 28 February 2016, 14:26:42 UTC | Imageio format: enforce public API: add some static keyword. | 28 February 2016, 14:26:42 UTC |
| 97a38ab | Roman Lebedev | 28 February 2016, 13:13:55 UTC | Views: enforce public API: add some static keyword. The idea is to export the same set of symbols with -fvisibility=hidden set, and without. | 28 February 2016, 13:13:55 UTC |
| 02f94ba | AlicVB | 26 February 2016, 20:29:45 UTC | masks: fix some leaks | 28 February 2016, 19:42:02 UTC |
| 6dadfb4 | Pedro Côrte-Real | 27 February 2016, 22:03:58 UTC | rawspeed: guard against another NULL-deref in MRW | 27 February 2016, 22:03:58 UTC |
| 52b9a8b | Pedro Côrte-Real | 27 February 2016, 21:40:30 UTC | cmake: set -pthread on C++ as well The new dt-rs-identify build would break on clang if this wasn't set. | 27 February 2016, 21:40:30 UTC |
| 467ef1f | Pedro Côrte-Real | 27 February 2016, 21:13:19 UTC | rawspeed: minor whitespace fixes | 27 February 2016, 21:13:19 UTC |
| 938a877 | Pedro Côrte-Real | 27 February 2016, 21:04:37 UTC | rawspeed: add missing cmake dependency identify needs to depend on rawspeed explicitely to make sure it actually gets rebuilt properly. | 27 February 2016, 21:04:37 UTC |
| 160dd91 | Pedro Côrte-Real | 27 February 2016, 20:06:56 UTC | rawspeed: add entry for Samsung GX20 DNGs Add entry to make the Samsung GX20 get proper naming everywhere. | 27 February 2016, 20:08:00 UTC |
| 7ccd19c | Ulrich Pegelow | 27 February 2016, 19:59:51 UTC | ashift: fix some comments | 27 February 2016, 19:59:51 UTC |
| 4578f2d | Pedro Côrte-Real | 27 February 2016, 19:38:46 UTC | rawspeed: cleanup cameras.xml finding in identify Isolate it into a function and make it work for any path size. | 27 February 2016, 19:40:32 UTC |
| c8d7fdb | Pedro Côrte-Real | 27 February 2016, 19:02:36 UTC | tools: improve make/model detection Instead of returning the rs-identify make/model directly still pass it through the exif->clean_name map to make sure we get the best possible name for the camera. We could just use the canonical_make/canonical_alias from identify but that may not be using the same cameras.xml file. | 27 February 2016, 19:40:32 UTC |
| 0ad1c5c | Pedro Côrte-Real | 27 February 2016, 18:55:49 UTC | rawspeed: use full cameras.xml path if possible Finding cameras.xml as a relative path from argv only works if we've been called with a path to the executable. Use the compiled path as the default to work around this for most sane installs. | 27 February 2016, 19:40:32 UTC |
| c9857b8 | Roman Lebedev | 27 February 2016, 17:57:44 UTC | CMake: project() after cmake_minimum_required(). cmake_minimum_required() already implies cmake_policy() | 27 February 2016, 19:15:47 UTC |
| 9a36a59 | Roman Lebedev | 27 February 2016, 17:04:32 UTC | CMake: no need for metadata_dummy lib. Replace with a simple dep. Refs #10354 | 27 February 2016, 19:15:47 UTC |
| 4c2c520 | Roman Lebedev | 27 February 2016, 16:58:14 UTC | Doc: CMake: darktableconfig.xml validity check as a prereq to prefs.html | 27 February 2016, 19:15:47 UTC |
| c563c97 | Roman Lebedev | 27 February 2016, 16:55:53 UTC | CMake: add darktableconfig.xml validity check as a prereq to prefs. | 27 February 2016, 19:15:47 UTC |
| c8d0d6f | Roman Lebedev | 27 February 2016, 16:52:52 UTC | CMake: add custom target to drive preferences generation. Refs #10354 As per add_custom_command() documentation, this is the only way. Else we risk race conditions. | 27 February 2016, 19:15:47 UTC |
| 202761c | Roman Lebedev | 27 February 2016, 16:52:36 UTC | CMake: add custom target to drive metadata generation. Refs #10354 As per add_custom_command() documentation, this is the only way. Else we risk race conditions. | 27 February 2016, 19:15:47 UTC |
| ffe559e | Roman Lebedev | 27 February 2016, 16:24:04 UTC | Data: CMake: check darktableconfig.xml via add_custom_target() | 27 February 2016, 19:15:47 UTC |
| 5984153 | Roman Lebedev | 27 February 2016, 16:10:12 UTC | External: RawSpeed: CMake: call xmllint via add_custom_target() As a consequence, the XML will be scanned each time make is run, not only at the time cmake is run, so any issues will be detected instantly, and not after next time cmake reprocesses the project. | 27 February 2016, 19:15:47 UTC |
| d45fe8c | Ulrich Pegelow | 27 February 2016, 17:35:55 UTC | ashift: better fit in both directions - avoid falling into local minimum | 27 February 2016, 17:35:55 UTC |
| 6d093c0 | Pedro Côrte-Real | 27 February 2016, 17:20:34 UTC | rawspeed: don't link identify against all of dt identify doesn't need anything from the dt API anymore so don't link it with the rest of darktable. | 27 February 2016, 17:20:34 UTC |
| 36205de | Pedro Côrte-Real | 27 February 2016, 16:53:42 UTC | rawspeed: stop special casing OSX in identify Simplify the identify code considerably by not doing the OSX special cases and not requiring the darktable API at all. | 27 February 2016, 16:53:42 UTC |
| 2a7d7e8 | Pedro Côrte-Real | 27 February 2016, 16:33:27 UTC | rawspeed: always print out make/model in identify Even if the camera is not supported print out make and model so it's useful for identifying unsupported files. | 27 February 2016, 16:33:27 UTC |
| ca90ee3 | Pedro Côrte-Real | 27 February 2016, 15:46:29 UTC | tools: use rs-identify in regression_tests Make camera detection more robust by using rs-identify to figure out make and model when exiv2 hasn't been able to figure that out. | 27 February 2016, 15:46:29 UTC |
| 8bb741d | Pedro Côrte-Real | 27 February 2016, 14:06:58 UTC | rawspeed: detect compressed RAF formats For now compressed RAF is broken so just refuse to decode it instead of coming up with a broken image. (refs #10934) | 27 February 2016, 14:06:58 UTC |
| 3b96b0a | Pedro Côrte-Real | 27 February 2016, 13:10:50 UTC | Basic support for Fuji X-Pro2 Thanks to Scott Gilbertson for the samples. Compressed files are not working yet. (ref #10934) | 27 February 2016, 13:10:50 UTC |
| 039a48d | Roman Lebedev | 27 February 2016, 12:11:57 UTC | External: RawSpeed: CMake: invalid cameras.xml => FATAL_ERROR, not WARNING Now that xmllint no longer barks at that xml, to prevent those issues from being re-introduced, we can and should do that. (especially since we build RS with -Werror) Next step will be turning that execute_process() into compile-time add_custom_target() - so that xmllint will be run each make, not just at cmake time. | 27 February 2016, 12:11:57 UTC |
| 0eb4720 | Pedro Côrte-Real | 27 February 2016, 11:44:08 UTC | rawspeed: fix xmllint errors There were a few issues with the DTD validation, none serious: - The ID definition had a copy paste error - XML DTDs are bonkers and force element order - The Camera definition was requiring too much - There were two or three typos | 27 February 2016, 11:44:08 UTC |
| 079de91 | Roman Lebedev | 27 February 2016, 09:29:48 UTC | External: RawSpeed: CMake: check cameras.xml validity. Test fails. | 27 February 2016, 09:29:48 UTC |
| 2909787 | Roman Lebedev | 26 February 2016, 16:40:22 UTC | Merge pull request #1160 from LebedevRI/cmake-issues CMake issues - version_gen related | 26 February 2016, 16:40:22 UTC |
| 64b202d | Ulrich Pegelow | 26 February 2016, 07:50:57 UTC | ashift: broader fitting range | 26 February 2016, 07:50:57 UTC |
| 2012723 | Roman Lebedev | 25 February 2016, 20:43:49 UTC | CMake: better PROJECT_VERSION if export-subst did not happen | 25 February 2016, 20:43:49 UTC |
| 7fbddea | Roman Lebedev | 25 February 2016, 18:34:00 UTC | De-duplicate git version string deduction into one new tool | 25 February 2016, 20:32:41 UTC |
| dc2dfcd | Roman Lebedev | 25 February 2016, 20:32:33 UTC | Fix cmake shenanigans with version_gen.h. Refs. #10354 E.g., previously, if you would do the following, the build would fail: $ cd ~/darktable/build/ && rm -rf * && CC=gcc CXX=g++ cmake ../ && make -j1 darkroom ... [ 38%] Building C object src/CMakeFiles/lib_darktable.dir/common/darktable.c.o In file included from /home/lebedevri/darktable/src/common/darktable.c:24:0: /home/lebedevri/darktable/build/src/version.h:7:25: fatal error: version_gen.h: No such file or directory compilation terminated. src/CMakeFiles/lib_darktable.dir/build.make:292: recipe for target 'src/CMakeFiles/lib_darktable.dir/common/darktable.c.o' failed make[3]: *** [src/CMakeFiles/lib_darktable.dir/common/darktable.c.o] Error 1 CMakeFiles/Makefile2:1290: recipe for target 'src/CMakeFiles/lib_darktable.dir/all' failed make[2]: *** [src/CMakeFiles/lib_darktable.dir/all] Error 2 CMakeFiles/Makefile2:1514: recipe for target 'src/views/CMakeFiles/darkroom.dir/rule' failed make[1]: *** [src/views/CMakeFiles/darkroom.dir/rule] Error 2 Makefile:626: recipe for target 'darkroom' failed make: *** [darkroom] Error 2 | 25 February 2016, 20:32:33 UTC |
| eaca247 | Roman Lebedev | 25 February 2016, 13:00:02 UTC | Move module_api.h into common/ <houz> LebedevRI: using a path for includes helps when some day another file with the same name pops up | 25 February 2016, 13:18:48 UTC |
| 6732b33 | Roman Lebedev | 25 February 2016, 09:09:49 UTC | Merge pull request #1159 from LebedevRI/api-headers Modules: only export those symbols which are listed in API headers | 25 February 2016, 09:09:49 UTC |
| 690ff6d | Roman Lebedev | 25 February 2016, 08:51:24 UTC | Iops: only export the symbols listed as public api. | 25 February 2016, 09:03:28 UTC |
| 303453b | Roman Lebedev | 24 February 2016, 17:05:46 UTC | Iops: add introspection functions into API header. | 25 February 2016, 09:03:28 UTC |
| a90e964 | Roman Lebedev | 25 February 2016, 08:51:02 UTC | Libs: only export the symbols listed as public api. | 25 February 2016, 09:03:28 UTC |
| 5b2305c | Roman Lebedev | 25 February 2016, 08:50:50 UTC | Imageio storage: only export the symbols listed as public api. | 25 February 2016, 09:03:27 UTC |
| 30a61ae | Roman Lebedev | 25 February 2016, 08:50:35 UTC | Imageio format: only export the symbols listed as public api. | 25 February 2016, 09:03:27 UTC |
| 0d0946a | Roman Lebedev | 25 February 2016, 08:57:45 UTC | Add a header with dt_module_dt_version()/dt_module_mod_version() Fixes loading of views. | 25 February 2016, 09:03:27 UTC |
| ba567c4 | Roman Lebedev | 25 February 2016, 09:03:21 UTC | Views: only export the symbols listed as public api. | 25 February 2016, 09:03:21 UTC |
| 8a9488a | Roman Lebedev | 24 February 2016, 17:34:49 UTC | Fix dt_iop_clip_and_zoom_demosaic_passthrough_monochrome_plain() :) | 25 February 2016, 08:47:55 UTC |
| 3361f5d | Roman Lebedev | 24 February 2016, 16:47:50 UTC | dt_imageio_export_with_flags(): do not crash if there are no iops/no finalscale iop | 25 February 2016, 08:47:55 UTC |
| 317cb92 | Roman Lebedev | 24 February 2016, 14:20:02 UTC | Darktable: dt_init(): quit if loading of darkroom view failed. | 25 February 2016, 08:47:55 UTC |
| 188b2c4 | Dan Torop | 24 February 2016, 23:20:54 UTC | x-trans: back to offsets as negative guards This reverts commit cac2a5c85876a5f8cb1c80bca2fbfca9dc3a537a, and uses a larger offset as negative guard. To be safe, asserts are also used. In addition, hexmap is now a function rather than a macro, and its results are properly set as const in all cases. | 24 February 2016, 23:42:24 UTC |
| 1b06167 | Roman Lebedev | 24 February 2016, 12:35:47 UTC | create_version_h.sh: properly handle filenames with spaces $ ninja [1/1] Updating version string /home/lebedevri/darktable/tools/create_version_h.sh: 18: [: /home/lebedevri/darktable/odd: unexpected operator | 24 February 2016, 12:35:47 UTC |
| 5b70c4f | Roman Lebedev | 24 February 2016, 12:13:14 UTC | Use config_file() instead of execute_process() to copy darktableconfig.dtd. Resolves #1052 ... in the build directory. This lets CMake know that darktableconfig.dtd is a generated file. When Ninja is used as the build tool, this fixes the following warning: CMake Warning (dev): Policy CMP0058 is not set: Ninja requires custom command byproducts to be explicit. Run "cmake --help-policy CMP0058" for policy details. Use the cmake_policy command to set the policy and suppress this warning. This project specifies custom command DEPENDS on files in the build tree that are not specified as the OUTPUT or BYPRODUCTS of any add_custom_command or add_custom_target: data/darktableconfig.dtd | 24 February 2016, 12:13:14 UTC |
| 7f41ab2 | Roman Lebedev | 23 February 2016, 19:48:39 UTC | Merge pull request #1154 from LebedevRI/iop-dither Dither IOP: add non-SSE codepath | 23 February 2016, 19:48:39 UTC |
| cac2a5c | Dan Torop | 23 February 2016, 18:34:20 UTC | x-trans: generalize handling negative CFA offsets Up until now FCxtrans() guarded against a negative offset into the CFA by adding a constant to incoming row & col, with the constant based on knowledge of the most negative row & col used in demosaic algorithms. But this could cause a bug if a future use of FCxtrans() involves unexpectedly negative inputs. This commit defines a modulo function that always returns a positive value (so long it is given a positive divisor). Both FCxtrans() and the HEXMAP() macro use this function. The downside is that this change results in approx. 4% slower processing of an X-Trans image when using a minimal set of iops. | 23 February 2016, 18:49:45 UTC |
| f43bf88 | Tobias Ellinghaus | 23 February 2016, 17:34:19 UTC | Add BRG color profile for testing + small cleanup Having BRG as a known-broken profile helps in testing if color management is set up correctly. Also make some more functions static. | 23 February 2016, 17:34:19 UTC |
| 2378635 | houz | 23 February 2016, 13:51:07 UTC | Merge pull request #1158 from sbraitbart/master Updated the Hebrew translation and fixed the semicolon issue | 23 February 2016, 13:51:07 UTC |
| cccbe1d | Shlomi | 23 February 2016, 13:10:02 UTC | Updated Hebrew translation - fix missing semicolon | 23 February 2016, 13:10:02 UTC |
| da117e7 | Shlomi | 23 February 2016, 11:37:05 UTC | Updated Hebrew translation plus minor fixes | 23 February 2016, 11:37:05 UTC |
| 954b4f6 | Roman Lebedev | 23 February 2016, 09:28:46 UTC | Tiling: fix build without OpenCL | 23 February 2016, 09:28:46 UTC |
| ab2de61 | Richard Levitte | 22 February 2016, 23:46:48 UTC | [l10n] Partial swedish update | 22 February 2016, 23:46:48 UTC |
| d14cf87 | Dan Torop | 22 February 2016, 20:35:44 UTC | imageop_math.h: FCxtrans() back to int parameters In this case, incoming row and col could be negative. And as pedrocr points out, internal pipe structures general use int. | 22 February 2016, 20:35:44 UTC |
| a0feb35 | Ulrich Pegelow | 22 February 2016, 19:24:11 UTC | opencl: apply changes of PR1123 to OpenCL code path | 22 February 2016, 19:24:11 UTC |
| 75cc938 | Ulrich Pegelow | 22 February 2016, 18:04:48 UTC | opencl: apply changes of PR869 to OpenCL codepath (finally) | 22 February 2016, 18:04:48 UTC |
| 1ba444e | Dan Torop | 15 February 2016, 04:34:46 UTC | demosaic: macro lookup for allhex A local equivalent to FCxtrans(). Makes it easier to handle modulo and offsets to ensure non-negative values. | 22 February 2016, 17:34:40 UTC |
| 78b40d0 | Dan Torop | 15 February 2016, 04:57:54 UTC | imageop_math.h: no offset in fcol() for x-trans Don't need it as FCxtrans() provides the necessary offset. | 22 February 2016, 17:34:40 UTC |
| bb294bf | Dan Torop | 15 February 2016, 04:46:46 UTC | demosaic & imageop_math: increase FCxtrans offset Increase from 6 to 18 (must stay a multiple of 6) so that don't have to use additional harcoded offsets in CPU-based Markesteijn demosaic. So long as there's an addition already in FCxtrans(), this shouldn't alter performance, so might as well keep the offset in one place. An offset of 18 is enough to handle the most negative values from Markesteijn demosaic, and more than enough for VNG and the non-demosaic iops. | 22 February 2016, 17:34:40 UTC |
| 6ca063a | Dan Torop | 15 February 2016, 00:00:15 UTC | demosaic: take advantage of new FCxtrans() code Now that FCxtrans can take a roi and already has an offset of 6, can simplify Markesteijn demosaic code accordingly. Note that offset for mirroring a border pixel is reduced from 24 to 12, as row/col will always be >= -17, and there is already a +6 offset in FCxtrans(). It had been set unnecessarily high. | 22 February 2016, 17:34:40 UTC |
| 3470a29 | Roman Lebedev | 22 February 2016, 17:14:34 UTC | Merge pull request #1149 from AlicVB/mask_loop masks : avoid infinite imbrication fixes #10921 | 22 February 2016, 17:14:34 UTC |
| 2ff5bed | Tobias Ellinghaus | 22 February 2016, 11:37:15 UTC | Support presets in "more modules" This allows people to define sets of shown or favourited modules that they typically use for a given task and quickly switch between those sets. | 22 February 2016, 11:37:15 UTC |
