May 11, 2015, 02:42 (GMT) |
Merge branch 'master' into ui-preview-buttons |
May 11, 2015, 01:58 (GMT) |
Merge branch 'master' into ui-preview-buttons |
May 10, 2015, 21:55 (GMT) |
Tweak docstrings also including full module path here is a bit odd since its an internal module. |
May 10, 2015, 21:44 (GMT) |
tweaks to Python scripts |
May 10, 2015, 21:28 (GMT) |
Merge branch 'master' into ui-preview-buttons |
May 10, 2015, 10:57 (GMT) |
improving path handling in py samples for ui-previews |
May 10, 2015, 10:34 (GMT) |
Add safety check in PRV_DEFERRED_DATA Disallow PRV_DEFERRED_DATA to be used when use_deferred is false |
May 10, 2015, 10:18 (GMT) |
changes for over-allocation use more conventional `pointer + 1` to access over alloc data. Was confusing having 2 byte trailing member of struct. |
May 10, 2015, 10:08 (GMT) |
Correct accidental sizeof void*/char mismatch |
May 9, 2015, 10:35 (GMT) |
Merge branch 'master' into ui-preview-buttons |
May 9, 2015, 10:33 (GMT) |
Move again GPU init code before UI_init one (since icons use OGL, I assue they need it initiated first). Oddly enough this did not seem to be an issue, but... |
May 9, 2015, 10:29 (GMT) |
Fix bad alpha - PreviewImage are assumed to be premultiplied... |
May 8, 2015, 20:05 (GMT) |
Fix was correct, but let's be a bit more generic/clear here... |
May 8, 2015, 19:50 (GMT) |
fixing crash with deferred previews |
May 8, 2015, 17:54 (GMT) |
Deferred loading for PreviewImage and icons based on them. This only works with thumbnails/filepath currently, but later we may extend it with other 'generated' ways... When requesting a PreviewImage from path thumbnail, only path (and thumb type) are stored in over-allocated fashion at end of PreviewImage struct. Then, calling `BKE_previewimg_ensure()` will take care of loading actual data if needed. This is only used from RNA currently, however, since we do not want to do that from main thread in draw icon code. For drawing code itself, tweaked system already existing for ID previews to also handle deferred (thumbnails) ones. Side Note: current job-based preview generation seems really bad to me: * It is not thread-safe - not at all! Reading **and writing** data from job thread (IDs, preview's rect, ...), without any protection of course. * It is badly used and adds a huge overweight (one job **per preview**!!!). This has to be completely rewriten imho, probably in a similar way as what is done for filebrowser previews in asset-experiments branch, i.e. using much lighter BLI_task, and being much stricter about not sharing memory between worker and main threads, but this is outside of scope for this patch. |
May 8, 2015, 17:54 (GMT) |
Cleanup: #define -> enum, and get rid of useless braces in case's. |
May 8, 2015, 17:54 (GMT) |
DNA; document how to ignore a struct |
May 8, 2015, 17:54 (GMT) |
Add missing TEXTEDIT_UPDATE option |
May 8, 2015, 17:54 (GMT) |
Cleanup: use r_* prefix for return args |
May 8, 2015, 17:54 (GMT) |
Fix T44634: Slide edge not responding to ALT + WHEEL to change reference edge in "even mode". Modal events (TFM_MODAL_EDGESLIDE_UP/_DOWN) were eaten by NOP generic transform event handling... |
|