| 309cf26 | Pascal Obry | 30 December 2016, 10:00:57 UTC | undo: add support for masks. Masks are now supported by the undo/redo circuitry as done for the development data. | 16 January 2017, 21:54:23 UTC |
| 5e3703d | Pascal Obry | 30 December 2016, 09:41:58 UTC | undo: rework the implementation. The implementation of the undo/redo was simple but was letting the complexity in the hand of the API users. In fact there was a need to keep a prev state and and a current state, and to record only previous states and not the current one. This was mostly working (except for a redo bug discoverred while working on this new implementation) but was tedious. The new undo implementation is more complex but makes it far easier to use. That is, modules can just record the current state and they don't need to keep a previous state at all. The undo history and map support has been reimplemented with this new semantics and have been cleaned-up as a consequence. This work has been highly motivated by the fact that we want to support undo/redo for masks and that it was far too complex to do right with the previous implementation. | 16 January 2017, 21:54:23 UTC |
| d8e7d14 | Pascal Obry | 27 December 2016, 17:48:49 UTC | Remove large duplicated code into the mask support. The circuitry to write the masks into the database was duplicated in two functions. This commit introduce a new routine _write_form_db() to share the common code. | 16 January 2017, 21:54:23 UTC |
| 0450f24 | Pascal Obry | 28 December 2016, 10:17:08 UTC | undo: add support for tag to identify an undo record. This make it possible for a user of the undo circuitry to avoid some recording as undo data with the same tag won't be recorded. It is used by the history to avoid too many recording when dragging a curve. | 16 January 2017, 21:54:23 UTC |
| e7b101c | Pascal Obry | 28 December 2016, 10:15:27 UTC | history: initialize end data for undo circuitry. | 16 January 2017, 21:54:23 UTC |
| f988ef2 | Pascal Obry | 27 December 2016, 17:35:47 UTC | Add undo group into the map view. Now adding multiple images into the map can be reverted by a single undo ctrl-z. | 16 January 2017, 21:54:23 UTC |
| c40df26 | Pascal Obry | 27 December 2016, 17:32:49 UTC | Add support for undo group. We add support for undo group into the undo/redo support. This will be used to create atomic changes for a set of images. For example adding many picture into the map should be reverted by a single ctrl-z. | 16 January 2017, 21:54:23 UTC |
| 9e55f99 | Pascal Obry | 16 January 2017, 21:39:07 UTC | Add missing lua as debug option in the help message. | 16 January 2017, 21:40:05 UTC |
| 48fae49 | Pascal Obry | 16 January 2017, 20:12:55 UTC | When copying ensure that the list of iop are reset to null. This fixes "copy-parts" -> "copy" -> "paste". That is after a full copy if the set of iop is not reset the paste is partial. Closes #11461. | 16 January 2017, 20:14:22 UTC |
| 8b0a8e7 | Roman Lebedev | 16 January 2017, 15:11:47 UTC | Rawspeed submodule update: large rework, better code. Fixes #11398 | 16 January 2017, 15:19:01 UTC |
| 235209e | Tobias Ellinghaus | 16 January 2017, 12:45:21 UTC | Don't write XMP when they didn't change When the XMP already exists on disk then we hash it first, compare that hash to the final data we are about to write and skip writing if the hashes match. Unfortunately the hashes can't be cached in db as one major use case is using the same images with several dbs or changing the XMPs from the outside. Fixes #11421 | 16 January 2017, 12:45:21 UTC |
| 6917bee | Roman Lebedev | 15 January 2017, 18:02:08 UTC | Rawprepare iop: support HDRMERGE fp unnormalized DNG's. Fixes #10177. Let's hope that we never encounter normalized fp dng's :) | 15 January 2017, 19:15:25 UTC |
| d67931d | Roman Lebedev | 15 January 2017, 19:14:28 UTC | Noiseprofile schema: adapt for Leica | 15 January 2017, 19:14:28 UTC |
| 6b22ba9 | Roman Lebedev | 15 January 2017, 19:10:27 UTC | Merge remote-tracking branch 'upstream/pr/1416' into HEAD * upstream/pr/1416: noise profile: add Leica X2, fixes #10950 | 15 January 2017, 19:10:27 UTC |
| 650457a | Roman Lebedev | 15 January 2017, 16:23:34 UTC | Rawspeed submodule update: mainly hdrmerge dng support | 15 January 2017, 19:09:35 UTC |
| 2de9ffb | Roman Lebedev | 15 January 2017, 18:54:02 UTC | Dockerfile: ZLIB [ci skip] | 15 January 2017, 18:54:16 UTC |
| 4b67f41 | Roman Lebedev | 14 January 2017, 13:15:53 UTC | Rawspeed submodule update: BINARY_PACKAGE_BUILD fix. | 14 January 2017, 13:15:53 UTC |
| 6f6b4e0 | johannes hanika | 13 January 2017, 22:03:20 UTC | local laplacian: add compile guard for non-sse builds | 13 January 2017, 22:03:20 UTC |
| d2d1675 | Stefan Schöfegger | 13 January 2017, 19:25:41 UTC | noise profile: add Leica X2, fixes #10950 | 13 January 2017, 19:25:41 UTC |
| f870c45 | Roman Lebedev | 13 January 2017, 18:45:07 UTC | Merge pull request #1415 from schenlap/noise_fuji_x_m1 noise profile: add Fujifilm X-M1, fixes #11308 | 13 January 2017, 18:45:07 UTC |
| 7ff4192 | Stefan Schöfegger | 08 January 2017, 18:45:59 UTC | noise profile: add Fujifilm X-M1, fixes #11308 Thanks to Denis Saint-Marcoux for the data. | 13 January 2017, 18:31:30 UTC |
| d8b8196 | Roman Lebedev | 13 January 2017, 16:45:41 UTC | Merge remote-tracking branch 'upstream/pr/1414' * upstream/pr/1414: Sony ILCE-6500: Add basic camera support, fixes #11441 | 13 January 2017, 16:45:41 UTC |
| a2b94fd | Roman Lebedev | 13 January 2017, 16:17:02 UTC | Rawspeed submodule update: Sony ILCE-6500, rstest changes | 13 January 2017, 16:45:31 UTC |
| e0604a7 | Roman Lebedev | 12 January 2017, 14:06:38 UTC | CMake: explicitly link against math library. Maybe helps with OBS build failures? | 13 January 2017, 16:45:31 UTC |
| 0d46c2a | Stefan Schöfegger | 12 January 2017, 19:10:08 UTC | Sony ILCE-6500: Add basic camera support, fixes #11441 | 12 January 2017, 19:10:08 UTC |
| 9147b4a | parafin | 12 January 2017, 15:10:25 UTC | remove unmaintained Gentoo ebuild | 12 January 2017, 15:10:25 UTC |
| 06d0f4b | parafin | 12 January 2017, 10:08:58 UTC | Mac: display an error when too old OS version is detected | 12 January 2017, 10:08:58 UTC |
| 67a862d | Roman Lebedev | 11 January 2017, 20:55:37 UTC | Document git submodule init. Closes #11446 [ci skip] | 11 January 2017, 21:00:34 UTC |
| 5b9ec76 | Tobias Ellinghaus | 11 January 2017, 19:26:33 UTC | Revert change to official OpenCL header | 11 January 2017, 19:26:33 UTC |
| ef45705 | peterbud | 11 January 2017, 18:50:14 UTC | More manual picking from PR 1327 So far only code changes and nothing about the build system. | 11 January 2017, 18:50:14 UTC |
| 8808158 | Roman Lebedev | 11 January 2017, 17:14:16 UTC | create_release.sh: some magic to support submodules | 11 January 2017, 18:33:19 UTC |
| 8fc1bd2 | Roman Lebedev | 10 January 2017, 21:28:23 UTC | Rawspeed submodule update. Mostly cleanup. | 10 January 2017, 21:28:23 UTC |
| 928886e | Roman Lebedev | 10 January 2017, 18:03:13 UTC | Bump max iso limit for presets from 51200 to FLT_MAX. Fixes #11410 | 10 January 2017, 18:56:57 UTC |
| 900ef77 | Tobias Ellinghaus | 10 January 2017, 10:02:12 UTC | Block keyboard accels when editing camera controls Also rename some variables. Fixes #11442 | 10 January 2017, 10:02:12 UTC |
| d87ce67 | Roman Lebedev | 10 January 2017, 07:44:18 UTC | Merge remote-tracking branch 'upstream/pr/1411' * upstream/pr/1411: imageop_math: fix comment in x-trans downscale demosaic: do full demosaic for previews demosaic: more refactoring of quality code demosaic: refactor calls to get_quality() demosaic: more refactoring of quality code demosaic: better describe quality flags | 10 January 2017, 07:44:18 UTC |
| e228f58 | Roman Lebedev | 10 January 2017, 07:43:58 UTC | Merge remote-tracking branch 'upstream/pr/1413' * upstream/pr/1413: slider popup: grab only keyboard events, fixes #11393 | 10 January 2017, 07:43:58 UTC |
| 9dd6989 | Roman Lebedev | 10 January 2017, 07:43:29 UTC | Rawspeed submodule update: win build issue | 10 January 2017, 07:43:29 UTC |
| 1d15804 | Dan Torop | 10 January 2017, 04:05:39 UTC | imageop_math: fix comment in x-trans downscale Fix wording/indentation of a comment which makes the debatable claim that CFA lookups wouldn't be necessary if downscaling an X-Trans image by 1/6 or less. | 10 January 2017, 06:34:59 UTC |
| 7ca7c13 | Dan Torop | 10 January 2017, 05:24:14 UTC | demosaic: do full demosaic for previews Fix blurry previews: Remove exception for previews from full scale demosaic. This was an erroneous attempt to return to pre-PR1263 behavior, when previews were downscaled in the demosaic iop. As they are now downscaled before this (in _init_f()), they need to be properly demosaiced. | 10 January 2017, 05:58:44 UTC |
| 5b22a57 | Dan Torop | 09 January 2017, 04:29:03 UTC | demosaic: more refactoring of quality code Remove last get_quality() call outside of demosaic_qual_flags(), and add DEMOSAIC_MEDIUM_QUAL flag for case when are in full pixelpipe not at 1:1 and config allows a lower quality demosaic. Question: why is there not similar code in the OpenCL path? | 10 January 2017, 05:51:58 UTC |
| 9e73157 | Dan Torop | 09 January 2017, 04:15:26 UTC | demosaic: refactor calls to get_quality() Make this work happen in demosaic_qual_flags() | 10 January 2017, 05:51:58 UTC |
| 85b390b | Dan Torop | 09 January 2017, 04:09:43 UTC | demosaic: more refactoring of quality code Make process_default_cl() and process_cl() use demosaic_qual_flags(). | 10 January 2017, 05:51:58 UTC |
| ebc71cd | Dan Torop | 09 January 2017, 03:56:05 UTC | demosaic: better describe quality flags Be more specific, and refer to "qual_flags" rather than "op_flags". Refactor the quality decision code to make clear that it depends on the image not being a preview. | 10 January 2017, 05:51:58 UTC |
| 78e6b48 | Roman Lebedev | 09 January 2017, 21:05:35 UTC | Merge remote-tracking branch 'upstream/pr/1412' * upstream/pr/1412: Leica D-Lux (Typ 109): Add basic camera support, fixes #11288 | 09 January 2017, 21:05:35 UTC |
| bef01b8 | Roman Lebedev | 09 January 2017, 21:05:15 UTC | Rawspeed submodule update > Leica D-Lux (Typ 109): add basic camera support .... | 09 January 2017, 21:05:15 UTC |
| c2e30fb | Christian Stussak | 09 January 2017, 20:35:46 UTC | slider popup: grab only keyboard events, fixes #11393 After the introduction of a separate code path for GTK 3.20+, numerical input for sliders was broken on some platforms that used this code path. The old code path just grabbed for keyboard events, while the GTK 3.20+ code path grabbed all kinds of events. This commit adjusts the GTK 3.20+ code accordingly. Only keyboard events are grabbed and owner_events is set to FALSE. This fixes problems with numerical input to the slider popup on macOS and possibly other platforms where GTK 3.20+ is used. | 09 January 2017, 20:35:46 UTC |
| bb47089 | Stefan Schöfegger | 08 January 2017, 19:53:59 UTC | Leica D-Lux (Typ 109): Add basic camera support, fixes #11288 | 09 January 2017, 18:44:00 UTC |
| 65a0bac | Ulrich Pegelow | 09 January 2017, 18:24:17 UTC | lens: minor cleanup | 09 January 2017, 18:24:17 UTC |
| 6ae2873 | Tobias Ellinghaus | 09 January 2017, 16:04:49 UTC | Properly delete duplicate XMP When removing images from db we rewrite the sidecar file to make sure that the latest data is stored on disk. Thus we have to remove images before we delete the XMP from disk, otherwise dt will just recreate it. | 09 January 2017, 16:04:49 UTC |
| c093cce | Tobias Ellinghaus | 09 January 2017, 15:04:15 UTC | Overwrite Xmp rating from raw when exporting Surprisingly exiv2 seems to not set the Xmp value when it's present already. This can lead to raw files with rating set by the camera not having the rating set in dt in the exported files. We might have to do similar for other fields that could be set in the raw files. | 09 January 2017, 15:04:15 UTC |
| 9d1cabd | Roman Lebedev | 08 January 2017, 18:39:36 UTC | Merge remote-tracking branch 'upstream/pr/1338' * upstream/pr/1338: lens: openmp style tweak lens: avoid fminf/fmaxf in modify_roi_in() lens: make better use of multi-processing in modify_roi_in() lens: add missing "for" in pragma omp parallel clause lens: optimized modify_roi_in() which is more friendly to the tiling mechanism lens: re-enable tiling | 08 January 2017, 18:39:36 UTC |
| ba5f891 | Roman Lebedev | 08 January 2017, 17:57:23 UTC | Merge remote-tracking branch 'upstream/pr/1410' * upstream/pr/1410: ignore UTF-8 BOM on gpx read, fixes #11435 | 08 January 2017, 17:57:23 UTC |
| bd7416a | Ulrich Pegelow | 08 January 2017, 17:24:29 UTC | lens: openmp style tweak | 08 January 2017, 17:24:29 UTC |
| 0a59fc7 | Ger Siemerink | 08 January 2017, 15:25:52 UTC | Updated Dutch translation | 08 January 2017, 15:25:52 UTC |
| cd6faaa | Ger Siemerink | 08 January 2017, 15:01:42 UTC | Updated Dutch translation | 08 January 2017, 15:01:42 UTC |
| 3246020 | Ger Siemerink | 08 January 2017, 14:57:07 UTC | Updated Dutch translation | 08 January 2017, 14:57:07 UTC |
| 26b7112 | Boucman | 08 January 2017, 10:00:19 UTC | Merge pull request #1384 from GHswitt/master Sync XMPs after tag attach/detach via Lua API. | 08 January 2017, 10:00:19 UTC |
| 6a2058b | Ulrich Pegelow | 07 January 2017, 21:43:58 UTC | lens: avoid fminf/fmaxf in modify_roi_in() | 07 January 2017, 21:43:58 UTC |
| d6522a0 | Roman Lebedev | 07 January 2017, 21:02:38 UTC | Update rawspeed submodule: less leaks | 07 January 2017, 21:05:11 UTC |
| 602de45 | johannes hanika | 07 January 2017, 19:26:39 UTC | base curve: allow multi instance and blending seems to make more sense this way | 07 January 2017, 19:26:39 UTC |
| d5a8c93 | Stefan Schöfegger | 07 January 2017, 16:47:33 UTC | ignore UTF-8 BOM on gpx read, fixes #11435 | 07 January 2017, 16:47:33 UTC |
| fd87ec6 | Boucman | 07 January 2017, 15:57:43 UTC | [lua] work around lua 5.3 not being compatible with lua 5.3 there is a compile-time option that change how ipairs work to be compatible with lua5.2 however this option is not forward compatible, so when it's activate, the code won't work with "real" lua 5.3 debian (at least) have activated that option as a workaround, reimplement ipairs to work the real 5.3 way | 07 January 2017, 15:58:05 UTC |
| 39b1463 | Ulrich Pegelow | 07 January 2017, 15:23:51 UTC | lens: make better use of multi-processing in modify_roi_in() | 07 January 2017, 15:23:51 UTC |
| e802d67 | Roman Lebedev | 07 January 2017, 11:04:42 UTC | {Docker,Brew}file: add lua5.3 For docker, i left the 5.2 there, because else we'd need to have 2 images - for master, and for 2.2.x | 07 January 2017, 11:04:42 UTC |
| 4de5a79 | Boucman | 07 January 2017, 10:06:51 UTC | [lua] bump core to 5.3 | 07 January 2017, 10:07:51 UTC |
| 0269642 | Boucman | 07 January 2017, 09:40:30 UTC | [lua] pygy require is not compatible with 5.3 and is not needed anymore anyway | 07 January 2017, 10:07:51 UTC |
| a75794a | Boucman | 07 January 2017, 09:17:41 UTC | [lua] upgrade build system to 5.3 | 07 January 2017, 10:07:51 UTC |
| decb2c7 | Boucman | 07 January 2017, 09:11:18 UTC | upgrade internal lua to 5.3.3 this does not make DT lua 5.3compatible yet | 07 January 2017, 10:07:51 UTC |
| 5ab6eb8 | Ulrich Pegelow | 07 January 2017, 08:39:37 UTC | Merge pull request #1407 from VasylTretiakov/patch-1 Typo fix in crossprocessing.xml | 07 January 2017, 08:39:37 UTC |
| 91e832c | Ulrich Pegelow | 07 January 2017, 08:39:13 UTC | Merge pull request #1408 from VasylTretiakov/patch-2 Typo fix in velvia.xml | 07 January 2017, 08:39:13 UTC |
| a960588 | Roman Lebedev | 06 January 2017, 21:05:48 UTC | Add some basic docs about the submodule. [ci skip] | 06 January 2017, 21:10:48 UTC |
| 2e0d4a3 | Boucman | 06 January 2017, 20:36:06 UTC | [lua doc] fix documentation error | 06 January 2017, 20:36:44 UTC |
| 898ab0c | Boucman | 26 November 2016, 11:13:31 UTC | [lua] disable dt.control.read under windows | 06 January 2017, 20:36:44 UTC |
| 2d5bdb8 | Boucman | 16 October 2016, 09:34:22 UTC | lua: add dt.print_log, like print_error but without the ERROR prefix | 06 January 2017, 20:36:44 UTC |
| 3aacc20 | Boucman | 06 January 2017, 20:30:44 UTC | post release version bump | 06 January 2017, 20:36:44 UTC |
| c5a922c | Dan Torop | 01 January 2017, 00:22:02 UTC | imageop_math.c: increase filter size by sample footprint It appears 24MP X-Trans images need a larger filter size to avoid artifacts, and this is proportional to px_footprint. Note this is tested on 12MP, 16MP, and 24MP X-Trans images. There's no guarantee that this will work well if higher (or lower) resolution sensors with this CFA ever emerge. | 06 January 2017, 19:39:45 UTC |
| 6a1acd8 | Dan Torop | 31 December 2016, 17:41:03 UTC | imageop_math.c: fix float/int mistake in filter width | 06 January 2017, 19:39:45 UTC |
| 5d35e0f | Dan Torop | 31 December 2016, 14:54:26 UTC | imageop_math.c: slightly smaller box filter Can get away with 8x8-ish filter to anti-alias x-trans downsamples. Also don't mention potential optimizations. A ring buffer would need to keep a separate buffer for each channel, too much work for the results. And because the input sample position moves by px_footprint which is not a multiple of the CFA width, a lookup would be necessary for each input CFA position for each output channel, it seems like an elaborate implementation which makes less clear code only run once when an image is loaded in darkroom. | 06 January 2017, 19:39:45 UTC |
| 64b369c | Dan Torop | 31 December 2016, 03:26:10 UTC | imageop_math.c: fix x-trans float kernel size Was lopping off right/bottom edge of it. | 06 January 2017, 19:39:45 UTC |
| 928a561 | Dan Torop | 31 December 2016, 03:03:01 UTC | imageop_math.c: always 9x9 box filter for x-trans The box filter for anti-aliasing X-Trans images should always be 9x9 (minimum size for good results). Its size depends on the CFA of the incoming image, not on the scaling ratio. Because the filter dimensions are known, can remove some asserts. | 06 January 2017, 19:39:45 UTC |
| 2509a7c | Dan Torop | 27 December 2016, 17:35:09 UTC | imageop_math.c: uint16_t x-trans downscale cleanups As LebedevRI points out, no need to cast division operands to float if the result is integer. Also, clean up spacing and a conditional. | 06 January 2017, 19:39:45 UTC |
| 83dfdb3 | Dan Torop | 27 December 2016, 17:30:19 UTC | imageop_math.c: simplify/blur float x-trans downscale Simplify the dt_iop_clip_and_zoom_mosaic_third_size_xtrans_f() path as with dt_iop_clip_and_zoom_mosaic_third_size_xtrans(), by only calculating the relevant color and not constraining sampling to a 3x3 subset of the CFA grid. Blur to get rid of color artifacts at edges. Note that the prior code offset the samples by -0.5*px_footprint, but this produces a misalignment with the results of the full-processed path, hence I removed the offset. | 06 January 2017, 19:39:45 UTC |
| 551f29a | Dan Torop | 27 December 2016, 02:09:46 UTC | imageop_math: blur while downsample (x-trans) For the x-trans int path, do a simple blur while downsampling mosaiced images. This gets rid of color fringes (confetti). See #11167. | 06 January 2017, 19:39:45 UTC |
| d384432 | Tobias Ellinghaus | 06 January 2017, 18:54:50 UTC | Don't show db lock popup when dbus is used ... ... to pass images loaded from command line to the running instance. | 06 January 2017, 18:54:50 UTC |
| 250164f | Roman Lebedev | 06 January 2017, 17:33:09 UTC | Revert submodule part of "ashift: document required variable name change in header of ashift_lsd.c" This partially reverts commit 5d963fe085a452d4f186ed0f4047bb88ac687927. | 06 January 2017, 17:34:53 UTC |
| 9e06d76 | Roman Lebedev | 06 January 2017, 16:59:59 UTC | Merge remote-tracking branch 'upstream/pr/1382' * upstream/pr/1382: demosaic: demosaic_op_flags() cleanups/fixes demosaic: refactor operation flags into function demosaic: revert downscaling of MIP_Fs demosaic: downscale previews for OpenCL and tiling demosaic: only shrink mosaiced MIP_Fs demosaic: downscale MIP_Fs demosaic: clearer full_scale_demosaicing condition demosaic: hack to interpolate previews | 06 January 2017, 16:59:59 UTC |
| b510522 | Vasyl Tretiakov | 06 January 2017, 08:43:53 UTC | Typo fix in velvia.xml "Its effect is tailored to increases saturation" should be "Its effect is tailored to increase saturation" | 06 January 2017, 08:43:53 UTC |
| 2ed9fab | Vasyl Tretiakov | 06 January 2017, 08:39:51 UTC | Typo fix in crossprocessing.xml "Cross-processing is a analog processing technique" should be "Cross-processing is an analog processing technique" | 06 January 2017, 08:39:51 UTC |
| bcd2267 | Ulrich Pegelow | 06 January 2017, 08:12:16 UTC | Merge pull request #1367 from upegelow/mask_free_fix drawn masks: reconsider use of dt_masks_free_form() towards a fix for… | 06 January 2017, 08:12:16 UTC |
| a2e6e90 | Ulrich Pegelow | 06 January 2017, 06:25:01 UTC | drawn masks: use cast instead of clumsy alias function | 06 January 2017, 06:25:01 UTC |
| e86afe1 | Ulrich Pegelow | 05 January 2017, 18:53:43 UTC | masks: fix for #11429 use auto-growing buffer instead of relying on any pre-calculated fixed size | 05 January 2017, 18:53:43 UTC |
| 5d963fe | Ulrich Pegelow | 05 January 2017, 18:20:39 UTC | ashift: document required variable name change in header of ashift_lsd.c | 05 January 2017, 18:20:39 UTC |
| 41d52bf | Roman Lebedev | 05 January 2017, 17:36:40 UTC | Travis-ci: <url> | 05 January 2017, 17:38:14 UTC |
| 3e06aa3 | Roman Lebedev | 05 January 2017, 17:28:44 UTC | Submodule src/external/rawspeed 8c0a57825..15a843cbd: > Travis-ci: spam into ML and IRC > DngDecoder: enable ASSHOTWHITEXY wb field reading. > Add Nikon LS-5000. mode="dng". (scanner) > Add PENTAX K10D. mode="dng". > Add PENTAX K-r. mode="dng". > Add Nokia Lumia 1020 mode="dng". > Add PENTAX K-5 mode="dng". Fixes #11433 | 05 January 2017, 17:38:14 UTC |
| 724db81 | peterbud | 05 January 2017, 12:23:11 UTC | Some cleanups, picked from PR 1327 These are generic code changes not directly related to the PR's purpose, so let's pick them separately to make the changes to reason about less. | 05 January 2017, 12:23:11 UTC |
| 51292ba | Dan Torop | 05 January 2017, 04:41:39 UTC | demosaic: demosaic_op_flags() cleanups/fixes tiling_callback() used a different condition (roi_out->scale > 0.8) than the process paths (roi_out->scale > 0.667). They should share the same logic, though. Furthermore, I'd tried to keep the old behavior when I refactored the logic into demosaic_op_flags(), but not only does the old behavior seem wrong, but it wasn't even kept, as demosaic_op_flags() would take precedence. Also, for succinctness, combine multiple bitfield setting in demosaic_op_flags() to single assignments. | 05 January 2017, 05:23:57 UTC |
| ecbeb38 | Dan Torop | 05 January 2017, 04:25:34 UTC | demosaic: refactor operation flags into function Separate flags full_scale_demosaicing, only_vng_linear, and xtrans_full_markesteijn_demosaicing were calculated using the same (or very similar) logic in CPU and various OpenCL processing paths, plus in tiling_callback(). This pulls all this logic into a single function, and tries to make it more legible. Note that the resulting function may not be as efficient as strung together ||, but this should be a negligible slowdown. | 05 January 2017, 04:38:20 UTC |
| d498bbb | Dan Torop | 05 January 2017, 02:46:13 UTC | demosaic: revert downscaling of MIP_Fs This reverts the two commits which don't successfully force downscaling of mosaiced MIP_Fs. It still may be a good idea to downscale them, but it will take more knowledge/thought to do so. | 05 January 2017, 04:30:43 UTC |
| 8ff91a2 | Pascal Obry | 04 January 2017, 21:05:02 UTC | Update French translation. | 04 January 2017, 21:05:49 UTC |
| ef8d700 | johannes hanika | 04 January 2017, 19:25:41 UTC | local contrast: add tooltips | 04 January 2017, 19:25:41 UTC |
