Re: 2.3.1; traveling; request for bug help

[Owen Taylor <otaylor redhat com>]

OK, went through and commented / marked with my opinion for milestone
or resolution. I didn't usually actually move the bugs there ... leaving
that for you :-)

Many thanks,
						Owen

On Sun, 2003-12-07 at 18:15, Soeren Sandmann wrote:

> I went over the GDK bugs.

[...]

> 2.2.4, 2.2.5 Milestones:
> 
> -  5878	 Gtk+ spews warnings when you call xsetpointer
>    Hard to fix, and a misconfiguration issue.

[2.4.1]

> -  73047  Wrong colors on big-endian/strange-visual hardware
>    Needs to be investigated by someone with the right hardware

[2.4.1]

> ?  73423  GDK_ISO_Left_Tab
>    For XKB GDK should fudge the keymap so that Shift+Tab returns GDK_ISO_Left_Tab

[2.4.1]

> -  115389 Crash dragging over Opera
>    Needs to be investigated with someone who has Opera

[2.4.1]

> -  118621 If there is only a static libXinerama.a library, gdk doesn't link with it
>    Doesn't happen on normal XFree86 setups

INCOMPLETE

> *p 119722 gdk/Xft2 and bitmap drawables bug
>    Has a patch from Owen, confirmed to work by reporter

I've sucessfully forgotten what my reservations were on this. So, it's
time to apply it :-). [2.4.0]

> +  126212 crash in gdk_pixbuf_from_drawable()
>    A crash, so should be fixed if possible. The problem is that only
>    the reporter can reproduce it.

[2.4.0] but I dont' see how we are going to progress on this one; wait
for someone else to run into the same thing, I guess.

> 2.4 API Freeze Milestone:
> 
> +p 64613  Focus handling (OutputOnly, ICCCM compliance)
>    Has a patch to switch to globally active windows, that needs review. 
>    Showstopper for accessibility. (I won't pretend I understand the issues)

Simple patch is conceptually OK to commit, added some comments.

> -  75497  Add colormaps to bitmaps 
>    Some drawing operations need to interprete colors. As I understand it, the patch in
>    bug 119722 fixes the most important case.

[2.6 API Freeze]. May eventually WONTFIX depending on exactly how we 
do Cairo integration.

> +p 95897  Add gdk_spawn() functions
>    Has a patch that needs review.

Added conditional commit approval.

> ?p 99276  missing API: gdk_rgb_set_max_colors()
>    Adds API to make GdkRGB allocate fewer colors. The attached patch is trivial, but
>    doesn't in itself fix the flashing problem on 8bit displays. To do that you would
>    also need an environment variable or something that would make gtk+ limit 
>    the number of colors allocated

Patch isn't per-colormap, so not useful. Moved to 2.6 API freezeL

> -  105101 Cross-platform "tray icon" API
>    Clear punt. This is major task.

[2.6 API Freeze]

> -p 106751 Add gdk_region_set_rectangle()
>    gdk_region_new() shows up on some profiles, but there it isn't clear that this
>    patch is the right fix.

Suggest retitle to "optimize region creation" [2.6.0]

> -  112182 Allow gdk_add_client_message_filter() to add more than one filter per event

I'm fine with the change, needs patch if it is to make 2.4 
[2.4 API FreezeL]

> +p 119375 Add way to get default group leader window
>    Has patch that needs review.

Added ok-to-commit

> 2.4.0 Milestone
> 
> -  50372  Setting stipple to NULL results in X error
>    Not extremely important. Possible for 2.4 if a patch shows up.

[2.4.1]

> -  53353  Check results of all X calls
>    Handle the case where user has called gdk_error_trap_push() so that GDK won't
>    see X errors. This can result in crashes when gtk+ assumes the calls succeeds.

[2.4.1]

> +p 60857  Antiexpose memory "leak"
>    Antiexpose records keep a reference to their window. This only affects memory use
>    when applications exits. This is a WONTFIX in my opinion, because trying to 
>    release all memory on exit is futile anyway.

It's trivial enough to fix (just use g_object_add_weak_pointer, two
lines change) that I think it's worth doing on aesthetic grounds.
[2.4.0L]

> -  63820  Refactor and document gdkgeometry-x11.c
>    A big job, not user-visible.

[2.6.0]

> -  72058  Configuration file that maps between XSETTING names and gdk names
>    A big job and not critical

[2.6.0]

> -  73424  Use gdk_keymap_translate_keyboard_state() instead of XLookupString()
>    I don't understand this issue. From Owen's comments it sounds like a
>    would-be-nice-but-not-critical

Was resolved some time ago. Dup'ed.

> -  85780  Add support for extra virtual modifiers
>    Would be nice, but not critical

Pretty high priority for 2.6, but not something we can do for 2.4 at
this point [2.6 API Freeze]

> +  89109  Add no-debug way to do --gdk-debug=NOGRABS
>    The reporter wants this as a workaround for his tablet which doesn't work when 
>    someone grabs the pointer. Owen says this is bad reason, because gtk+ without
>    grabs is a broken gtk+. This is a WONTFIX in my opinion, because I don't see
>    any non-debug use of a no-grabs option

WONTFIX

> ?  97510  remove static gravity checks from GDK
>    Just needs a patch I guess, but non-critical

Yeah, it's an entirely trivial patch (make
gdk_window_set_static_gravities always return TRUE) [2.4.0]

> -  101362 Use XCopyArea() to scroll windows whose children are all input-only
>    A trivial performance optimization.

Not convinced it makes sense ... the last comment is an open question.
[2.6.0]

> ?  101983 gdk_draw_pixbuf() ignores the GC clip mask of the GC.
>    It looks moderately hard to fix, and the feature is rarely used, 
>    a "this simply doesn't work" bug.

Considering impending cairo, I'd say => docs.

> +  105497 const GdkColor 
>    Purely janitorial. Looks like 20 minutes of fun with grep.

[2.4.0L]

> ?  110224 Make xsettings-client use gdk_x11_display_grab()
>    Looks reasonably easy to fix. I'd say 2.4.0 if a patch shows up,
>    otherwise punt.

Moderately important, actually. [2.4.0]

> -  113310 Unset bg when mapping override-redirect toplevels
>    Cosmetic, and fairly hard to fix

I'd really like to see someone working on this one, but it's certainly
not something 2.4 should be blocking on. [2.6.0, maybe do a backport
later]

> -  113394 Action negotation for Motif DND
>    Fixing this involves protocol snooping two Motif applications talking
>    to eachother.

[2.6.0]

> -  113476 Use XftDrawSetClipRectangles()
>    This function is broken in the released version of Xft, so this bug can't
>    be fixed before a new version of Xft is out.

[2.6.0]

> ?p 115650 Support _NET_WM_USER_TIME
>    There is a patch that adds new API and an X11 implementation. It isn't
>    clear if we want to support the hint.

Definitely need to support the hint. Added comments to the bug about
what needs to be done to the patch. (Its quite incomplete). Going to
be hard to get this into 2.4, but let's leave it on 2.4 API Freeze 
for now.

> ?p 116626 Use keyboard map contents to detect RTL groups
>    This patch has C++ code that can easily be converted into C.
>    To me this looks easy to fix if the relevant Pango patch is applied, 
>    but Owen will have to say.

Yes, the relevant Pango patch is a must-fix for the 2.4 API freeze, 
so lets leave this open for 2.4.0.

> +  116719 Panel crashed
>    I am pretty sure this is a duplicate of 117579, but even if it isn't, it
>    is a crasher, and I think it should be fixed.
> 
> +  117579 _gdk_xid_table_insert trouble
>    Causes crashes when gdk_window_reparent() has been called. I think it is
>    important to fix this one.

Yep [2.4.0H]

> +  119804 Hide -lXcursor dependency
>    Must be fixed probably, because otherwise Xcursor will be in the ABI.

[2.4.0]

> +p 125645 GC caching
>    There is a patch that needs to be reviewed.

Added various comments. Lets leave on 2.4.0 for now, though it's not 
a blocker.

> "---" Milestone:
> 
> -  89179  Reporter wants a way to differentiate between focus-change events that
>    are the result of a grab operation, and regular focus-change events
>    Bug is NEEDINFO at the moment.

I bugged Mark for more details.

> -  121471 Window specific filter functions only sees events gtk+ knows about
>    One suggested fix here is making _gdk_x11_register_event_type() public.
>    This is not urgent, because you can use a global filter function, and 
>    check the window yourself.

OK for 2.4 if someone comes up with the trivial patch. But really low
priority. [2.4 API FreezeL]

> +p 122448 Make hooks around gdk_threads_enter/leave possible
>    This has been discussed on gtk-devel recently. The latest version of 
>    the patch needs to be reviewed.

Applied a couple of days ago.

> -  122728 No Busy+Interactive Cursor
>    Add a combined hourglass+arrow cursor. I don't think this is 2.4 material

[2.6 API freeze]

> ?  122777 Type-punned pointer warnings
>    Janitorial. Could be 2.4.0 material if a patch show up

We should do something here, if nothing else, -Wno-strict-aliasing
[2.4.0] 

> +p 123546 Workaround for broken tablet driver
>    This patch is a workaround for a broken driver. It looks reasonable to me.

Added a conditional OK-to-commit [2.2.5]

> ?p 126375 Way to detect other application owning selection
>    I am not sure I understand this bug at all. Owen asks what the point is
>    of having gdk API, rather than just using Xlib.

Marked WONTFIX

> -  127927 Build fails
>    Build fails for Kjartan, but not a showstopper

glib-genmarshal problem. Moved to gobject.

> ?p 128005 Gdk converts 32 bit coordinates into 16 bit coordinates
>    This is either really easy to deal with (NOTABUG), or extremely difficult (do
>    complex math to figure out what to convert 32 bit coordinates into).
>    In my opinion the complex math is not worth it, so this is NOTABUG.

Some simple easy to fix changes might make sense [2.6.0]

> ?p 128529 Can't enter UKRAINIAN_GHE_WITH_UPTURN
>    I don't know if this is easy to fix or not.

Added an ok-to-commit. [2.2.5]