libui/TODO.md at 98e06716b6b3ff029ac112b2759a15c8476d4021

16 KiB

Raw Blame History

make sure the last line of text layouts include leading
documentation notes:
- static binaries do not link system libraries, meaning apps still depend on shared GTK+, etc.
- ui*Buttons are NOT compatible with uiButton functions
more robust layout handling
- uiFormTie() for ensuring multiple uiForms have the same label area widths
- uiSizeGroup for size groups (GtkSizeGroup on GTK+, auto layout constraints on OS X; consider adding after 10.8 is gone)
windows: should the initial hwndInsertAfter be HWND_BOTTOM for what we want?
windows: document the rules for controls and containers
windows: document the minimum size change propagation system
provisions for controls that cannot be grown? especiailly for windows
LC_VERSION_MIN_MACOSX has the 10.11 SDK; see if we can knock it down to 10.8 too; OS version is fine
- apply the OS version stuff to the test program and examples too
- what about micro versions (10.8.x)? force 10.8.0?
go through ALL the objective-c objects we create and make sure we are using the proper ownership (alloc/init and new are owned by us, all class method constructors are autoreleased - thanks mikeash)
on OS X, edit shortcuts like command-C working require that the menu entries be defined, or so it seems, even for NSAlert
- other platforms?
make sure all OS X event handlers that use target-action set the action to NULL when the target is unset
provide a way to get the currently selected uiTab page? set?
make it so that the windows cntrols only register a resize if their new minimum size is larger than their current size to easen the effect of flicker
- it won't remove that outright, but it'll help
add an option to the test program to run page7b as an independent test in its own window
- same for page7c
http://blogs.msdn.com/b/oldnewthing/archive/2004/01/12/57833.aspx provide a DEF file on Windows
all ports: update state when adding a control to a parent
should uiWindowsSizing be computed against the window handle, not the parent?
DPI awareness on windows
http://stackoverflow.com/questions/4543087/applicationwillterminate-and-the-dock-but-wanting-to-cancel-this-action

ultimately:

MAYBE readd lifetime handling/destruction blocking
related? [12:25] And the blue outline on those buttons [ALL clicked buttons on Windows 7] won't go away
- I get this too
  - not anymore
SWP_NOCOPYBITS to avoid button redraw issues on Windows when not in tab, but only when making resize faster
secondary side alignment control in uiBox
Windows: don't abort if a cleanup function fails?
32-bit Mac OS X support (requires lots of code changes)
- change the build system to be more receptive to arch changes

notes to self

explicitly document label position at top-left corner
explicitly document that if number of radio buttons >= 1 there will always be a selection
mark that uiControlShow() on a uiWindow() will bring to front and give keyboard focus because of OS X
- make sure ShowWindow() is sufficient for zorder on Windows
document that you CAN use InsertAt functions to insert at the first invalid index, even if the array is empty
note that uiTabInsertAt() does NOT change the current tab page (it may change its index if inserting before the current page)
note that the default action for uiWindowOnClosing() is to return 0 (keep the window open)
note that uiInitOptions should be initialized to zero
explicitly document that uiCheckboxSetChecked() and uiEntrySetText() do not fire uiCheckboxOnToggled() and uiEntryOnChanged(), respectively
note that if a menu is requested on systems with menubars on windows but no menus are defined, the result is a blank menubar, with whatever that means left up to the OS to decide
note that handling of multiple consecutive separators in menus, leading separators in menus, and trailing separators in menus are all OS-defined
note that uiDrawMatrixInvert() does not change the matrix if it fails
note that the use of strings that are not strictly valid UTF-8 results in undefined behavior
test RTL
- automate RTL
now that stock items are deprecated, I have to maintain translations of the Cancel, Open, and Save buttons on GTK+ myself (thanks baedert in irc.gimp.net/#gtk+)
- either that or keep using stock items
http://blogs.msdn.com/b/oldnewthing/archive/2014/02/26/10503148.aspx
build optimizations
use http://www.appveyor.com/ to do Windows build CI since people want CI
consider just having the windows backend in C++
- consider having it all in C++

don't forget LONGTERMs as well

notes

http://blogs.msdn.com/b/oldnewthing/archive/2004/03/29/101121.aspx on accelerators
group and tab should act as if they have no child if the child is hidden on windows
a way to do recursive main loops
- how do we handle 0 returns from non-recursive uiMainStep() calls that aren't the main loop? (event handlers, for instance)
should repeated calls to uiMainStep() after uiQuit() return 0 reliably? this will be needed for non-recursive loops

http://stackoverflow.com/questions/38338426/meaning-of-ampersand-in-rc-files/38338841?noredirect=1#comment64093084_38338841

label shortcut keys

remove whining from source code

[01:41:47] Hi. does pango support "fgalpha". I see that foreground="112233xx" works ( alpha=xx ), but fgalpha is a no-op [01:52:29] pango_attr_foreground_alpha_new (32767) seems to be called in either case, but only the "foreground" attr works [01:56:09] lolek (lolek@ip-91-244-230-76.simant.pl) joined the channel [01:57:48] ok. seems like "foreground" is mandatory attr, 1. "foreground-without-alpha" + "alpha" works 2. "foreground-with-alpha" works. 3. "alpha" alone doesn [01:57:52] 't work [01:58:29] Is there a way to just specify alpha on the current foreground color ? [02:00:23] lolek (lolek@ip-91-244-230-76.simant.pl) left the channel [02:07:41] mjog (mjog@uniwide-pat-pool-129-94-8-98.gw.unsw.edu.au) left IRC (Quit: mjog) [02:08:10] seb128 (seb128@53542B83.cm-6-5a.dynamic.ziggo.nl) joined the channel [02:12:37] huh [02:12:41] what version of pango? [02:13:05] the latest . [02:15:00] 1.40.3 [02:20:46] I'll ahve to keep this in mind then, thanks [02:20:59] if only there was a cairo-specific attribute for alpha...

FONT LOADING

[00:10:08] andlabs: is there API yet to load from memory? last i checked i only found from file (which we use in builder). https://git.gnome.org/browse/gnome-builder/tree/libide/editor/ide-editor-map-bin.c#n115 [00:13:12] mrmcq2u_ (mrmcq2u@109.79.53.90) joined the channel [00:14:59] mrmcq2u (mrmcq2u@109.79.73.102) left IRC (Ping timeout: 181 seconds) [00:15:19] hergertme: no, which is why I was asking =P [00:15:30] I would have dug down if I could ensure at least something about the backends a GTK+ 3 program uses [00:15:39] on all platforms except windows and os x [00:16:11] to the best of my (partially outdated, given pace of foss) knowledge there isn't an api to load from memory [00:16:28] you can possibly make a tmpdir and put a temp file in there [00:16:52] and load that as your font dir in your FcConfig, so any PangoFontDescription would point to that one font, no matter what [00:17:18] (using the API layed out in that link) [00:18:18] dsr1014__ (dsr1014@c-73-72-102-18.hsd1.il.comcast.net) joined the channel [00:35:18] simukis_ (simukis@78-60-58-6.static.zebra.lt) left IRC (Quit: simukis_) [00:35:48] dreamon_ (dreamon@ppp-188-174-49-41.dynamic.mnet-online.de) joined the channel [00:40:09] samtoday_ (samtoday@114-198-116-132.dyn.iinet.net.au) joined the channel [00:40:32] mjog (mjog@120.18.225.46) joined the channel [00:40:38] hergertme: not necessarily fontconfig [00:40:45] it can be with ft2 or xft I guess [00:40:55] especially since I want the API NOT to make the font part of the font panel [00:42:07] what sort of deprecated code are you trying to support? [00:42:35] both of those are deprecated in pango fwiw [00:43:06] on Linux im pretty sure we use FC everywhere these days [00:44:46] (and gtk_widget_set_font_map() is how you get your custom font into a widget without affecting the global font lists, as layed out in that link) [00:49:14] vasaikar (vasaikar@125.16.97.121) joined the channel [00:50:14] karlt (karl@2400:e780:801:224:f121:e611:d139:e70e) left IRC (Client exited) [00:50:49] karlt (karl@2400:e780:801:224:f121:e611:d139:e70e) joined the channel [00:51:43] PioneerAxon (PioneerAxo@122.171.61.146) left IRC (Ping timeout: 180 seconds) [00:57:47] PioneerAxon (PioneerAxo@106.201.37.181) joined the channel [01:03:01] karlt (karl@2400:e780:801:224:f121:e611:d139:e70e) left IRC (Ping timeout: 181 seconds) [01:05:49] muhannad (muhannad@95.218.26.152) left IRC (Quit: muhannad) [01:07:51] hergertme: hm [01:07:54] all right, thanks [01:08:05] hergertme: fwiw right now my requirement is 3.10 [01:10:47] ah, well you'll probably be missing the neccesary font API on gtk_widget [01:11:04] but pango should be fine even back as far as https://developer.gnome.org/pango/1.28/PangoFcFontMap.html [01:11:56] good [01:12:04] because this is for custom drawing into a DrawingArea [01:14:12] presumably just create your PangoContext as normal, but call pango_context_set_font_map() with the map you've setup. now, the load a font from a file i dont think was added to FontConfig until later though (not sure what release) [01:15:53] FcConfigAppFontAddFile() <-- that API [01:16:30] great, and they don't say what version the API was added in teh docs function: ide_editor_map_bin_add()

Mouse ClickLock: do we need to do anything special? should we? https://msdn.microsoft.com/en-us/library/windows/desktop/ms724947(v=vs.85).aspx
consider a uiAnticipateDoubleClick() or uiDoubleClickTime() (for a uiQueueTimer()) or something: https://blogs.msdn.microsoft.com/oldnewthing/20041015-00/?p=37553
determine whether MSGF_USER is for and if it's correct for our uiArea message filter (if we have one)
source file encoding and MSVC compiler itself? https://stackoverflow.com/questions/20518040/how-can-i-get-the-directwrite-padwrite-sample-to-work
- also need to worry about object file and output encoding...
- this also names the author of the padwrite sample
OpenType features TODOs
- https://stackoverflow.com/questions/32545675/what-are-the-default-typography-settings-used-by-idwritetextlayout
- feature/shaping interaction rules for arabic: https://www.microsoft.com/typography/OpenTypeDev/arabic/intro.htm
- other stuff, mostly about UIs and what users expect to be able to set
add NXCOMPAT (DEP awareness) to the Windows builds
- and ASLR too? or is that not a linker setting

OS X: embedding an Info.plist into a binary directly https://www.objc.io/issues/6-build-tools/mach-o-executables/ TODO will this let Dictation work? TODO investigate ad-hoc codesigning

https://blogs.msdn.microsoft.com/oldnewthing/20040112-00/?p=41083 def files for decoration (I forget if I said this earlier)

TODO ClipCursor() stuff; probably not useful for libui but still https://blogs.msdn.microsoft.com/oldnewthing/20140102-00/?p=2183 https://blogs.msdn.microsoft.com/oldnewthing/20061117-03/?p=28973 https://msdn.microsoft.com/en-us/library/windows/desktop/ms648383(v=vs.85).aspx

https://cmake.org/Wiki/CMake_Useful_Variables set(CMAKE_SHARED_LINKER_FLAGS "-Wl,--no-undefined") On Unix systems, this will make linker report any unresolved symbols from object files (which is quite typical when you compile many targets in CMake projects, but do not bother with linking target dependencies in proper order). (I used to have something like this back when I used makefiles; did it convert in? I forget)

look into these for the os x port https://developer.apple.com/documentation/appkit/view_management/nseditor?language=objc https://developer.apple.com/documentation/appkit/view_management/nseditorregistration?language=objc

for future versions of the os x port https://developer.apple.com/documentation/appkit/nslayoutguide?language=objc and anchors https://developer.apple.com/documentation/appkit/nsuserinterfacecompression?language=objc https://developer.apple.com/documentation/appkit/nsuserinterfacecompressionoptions?language=objc though at some point we'll be able to use NSStackView and NSGridView directly, so...

run os x code with OBJC_DEBUG_MISSING_POOLS=YES and other OBJC_HELP=YES options
- turn off the autorelease pool to make sure we're not autoreleasing improperly

16 KiB Raw Blame History

16 KiB

Raw Blame History