Subject: Re: [E-devel] [Patch] ecore_ipc - remove potential risk in
ecore_ipc_shutdown
I found a problem this infinite loop case.
If server is deleted, then ECORE_IPC_EVENT_SERVER_DEL callback
function will be called in client side.
It will happen infinite loop in ecore_ipc_shutdown if
ecore_ipc_shutdown called in this ECORE_IPC_EVENT_SERVER_DEL callback
function.
For example,
server_del_handler =
ecore_event_handler_add(ECORE_IPC_EVENT_SERVER_DEL, _server_del_cb, NULL);
static Eina_Bool
_server_del_cb(void *data, int type, void *event)
{
ecore_ipc_shutdown();
return EINA_TRUE;
}
If server is deleted,
1. _ecore_ipc_event_server_del : svr->event_count++
2. _server_del_cb : ecore_ipc_shutdown called
3. ecore_ipc_shutdown : while (servers) ecore_ipc_server_del(eina_list_data_get(servers))
4. ecore_ipc_server_del : can't eina_list_remove(servers, svr) because event_count != 0
5. infinite loop
I think this while code is very dangerous whether user miss or not.
I modified EINA_LIST_FOREACH_SAFE instead of EINA_LIST_FOREACH refer
to ecore_con.
Please review this patch.
SVN revision: 67874
missing Logfn's. Add handler to free the mouse_move event when we're
done with it. Add a function to retrieve the 'last mouse button down
time' (needed for fixing surface move).
SVN revision: 67781
ecore_imf_context.c: In function ‘ecore_imf_context_event_callback_del’:
ecore_imf_context.c:1099:9: warning: ‘return’ with no value, in function returning non-void
ecore_imf_context.c:1109:26: warning: initialization discards qualifiers from pointer target type
SVN revision: 67585
Long time ago, in
http://www.mail-archive.com/enlightenment-devel@lists.sourceforge.net/msg32795.html
mail thread and IRC,
I talked with about problem of asynchronous event API such as
ecore_imf_context_commit_event_add,
ecore_imf_context_preedit_changed_event_add, so on.
In short, The problem is that key event and text_set APIs are processed
immediately, but commit event and preedit changed event is processed
asynchronously because those APIs add each event to ecore event queue.
To fix these problems, I've decided to create synchronous event APIs such
as ecore_imf_context_event_callback_add, del and call.
For considering compatibility, sync and async event callback functions are
used in xim and scim immodule.
SVN revision: 67290
Hello e people, i modified some comments to get less doxygen
warnings/errors.
Signed-Off-By: Guillaume Friloux <guillaume.friloux@asp64.com>
SVN revision: 67270
cannot depend on the order in which the interfaces are given to us, so
we cannot create the data_device until we have both the input device
and the data_device_manager. Fix that.
Add more dnd code (not complete yet because frankly, the way Wayland
handles dnd & selection code is crappy, stupid, and difficult to
implement).
SVN revision: 67247
divide each new functions to two part
which create ecore evas and create socket for upper layer.
socket: new and listen
plug: new and connect
SVN revision: 67238
Subject: [E-devel] [PATCH] fix build error of ecore_imf_xim when using --disable-xim option
Date: Mon, 9 Jan 2012 19:53:17 +0900
Hello, EFL developers.
When building ecore with --disable-xim option, build error occurs.
This patch will solve this build error problem.
Would you please review and apply in svn?
SVN revision: 67191
Add more work on dnd code (incomplete still).
NB: We will need to add a method to request non-argb surfaces. For
now, all surfaces are argb.
SVN revision: 67180
After we have set the new surface for the evas engine. Reason being:
The shell_surface_listener_configure event handles resizing the
ecore_evas. If that gets called Before we have set the new engine
surface, then bad crashes can happen.
SVN revision: 67161
Add a check for valid connection in ecore_xcb_icccm.c
NB: For some reason, XCB does not like ecore's timestamp which is
being passed in. Since all the calls to ecore_x_window_focus_at_time
are being passed the current timestamp anyway, just use
XCB_CURRENT_TIME until I have more available 'free time' to dig deeper
into this.
SVN revision: 67136
events).
Add big fat NB wrt potential 'hack' (works around screen garbage on
resize).
Add evas damage rectangle on resize.
Do surface_damage before surface_attach.
Prevent compositor crash on resize by checking width/height <= 0.
Don't do a resize of the ecore_evas unless the event matches.
SVN revision: 67013
Add variable to store what button was pressed (part of focus issue
fixes).
Fix focus handling (see comments about wayland pointer focus).
SVN revision: 66994
It is possible that several process want to show same canvas.
For support this, I changed Ecore buffer canvas to server(socket)
If some application want to show this canvas,
it create plug image object.
SVN revision: 66892
specific 'surface' and 'time'.
Add function (internal) to send a mouse_move (cleans up code).
Add event 'multi' stuff for mouse button down.
Fix handle_pointer_focus function to not send 'focus' events (handled
differently).
Fix handle_keyboard_focus function to actually send 'focus' to new
windows.
NB: This fixes most of the focus issues I was having earlier.
SVN revision: 66859
(_ecore_evas_mouse_move_process already feeds to evas).
Add variable to store the 'timestamp' of when mouse button gets
pressed (needed for 'moving' a surface).
SVN revision: 66808
Subject: [E-devel] [Patch][ecore_con] Fix return value checking
Date: Tue, 3 Jan 2012 10:01:17 +0900
Hi,
There is invalid return value checking of curl_multi_perform() in ecore_con_url.
Because CURLM_CALL_MULTI_PERFORM is also not CURLM_OK,
CURLM_CALL_MULTI_PERFORM checking should precede CURLM_OK checking.
It's simple fix. Please review this
--
BRs,
Kim.
SVN revision: 66771
post data to be corrupted (as it was never copied into the
ecore con url struct) or could cause crashes if the memory
pointed to became invalid.
SVN revision: 66619
actually functional yet).
Implement the data device manager.
Implement the data device itself.
Fixes for shutdown so we destroy everything correctly (data device,
device manager, shell, etc).
SVN revision: 66581
Add an actual 'frame' to ecore_evas_wayland. (just a boring rectangle frame w/ the title).
Add support for size_min/max/base/step setting.
Add support for title & name_class setting.
Add support for setting a cursor.
Add support for hiding an ecore_evas.
SVN revision: 66578
Subject: Re: [E-devel] [PATCH] scim-EFL immodule
I'd like to contribute EFL-scim immodule.
It will be useful to users who wants to input Chinese, Japanese, Korean,
and a variety of languages that SCIM provides.
When you run, you should set environment variable 'ECORE_IMF_MODULE'
# export ECORE_IMF_MODULE=scim
SVN revision: 66507
Subject: Re: [E-devel] [Patch][ecore_con_url] Add proxy & timeout set
API
There existed two problems when using ecore_con_url. We cannot set
proxy and set timeout.
So, I added two APIs for solving these problems.
Proxy can be set by setting libcurl option.
Timeout also can be set but It need to add handler for Linux alarm
signal. (Please refer:
http://curl.haxx.se/libcurl/c/curl_easy_setopt.html#CURLOPTTIMEOUT)
So I implemented it by using ecore timer.
SVN revision: 66414
Subject: [E-devel] [PATCH] support to get style attribute of XIM module
ATM, edje entry doesn't display the style of preedit string such as
underline, reverse, and Highlight.
It's because XIM module doesn't provide the style attribute.
This patch is made for providing the style attibute information to
entry or other editable widget.
SVN revision: 66381
my testing, but documentation for g_static_mutex's isnt that great to
let us know what to do with a static mutex... considering it's...
static... pending respons from reporter for backport
SVN revision: 66360
Subject: [E-devel] [Patch] Add ecore_x API for error handling
I added ecore_x_error_resource_id_get() api. It returns the resource
id that caused the last X error. I think it will be useful for handling
X error.
SVN revision: 66136
animator keeps the animator handler in an infinite loop. Do
the same as timers and mark them as "just added" to skip in
this run of animators
SVN revision: 66109
I've also set it to abort() with a big error if it gets to the old fallthrough spot, so make sure to send me a backtrace if this is reached (though it should be impossible)
SVN revision: 66068
in general, this should provide an improvement which scales with the amount of data being transferred:
* small transfers will incur a small amount of overhead from potentially unneeded memory as I try to account for a bug in FIONREAD which returns a number that is smaller than the actual number of bytes available for read on a socket
* large transfers will no longer require any copies of the data
on systems which do not provide the FIONREAD ioctl(), old functionality will be used
this should work on windows, though I (obviously) can't test it myself
thus ends the longest commit message I have ever written
SVN revision: 66063
ECORE-CON-SOCKS! SOCKS ON!!!!
now ecore_con supports socks (v4 and v4a only, so no ipv6) connections natively for making remote connections
for those of you who want their apps to start proxying immediately, just update and export this handy environment variable:
ECORE_CON_SOCKS_V4=[user@]PROXY_IP_ADDRESS:PROXY_PORT[:1] <--use :1 here to enable dns lookups on the proxy
SVN revision: 65934
in later 2011, we began using formatting tools to fix problematic formatting caused by previous runs of formatting tools used to fix problematic formatting
SVN revision: 65852
elementary_config does this, and the glib based main loop didn't support it.
Signed-off-by: Mike McCormack <mj.mccormack@samsung.com>
SVN revision: 65683
* fix some event values
* Add the "Windows" key support
* manage modifiers for shift, ctrl, alt and windows keys
that support is not finished, there still several bugs.
SVN revision: 65613
A previous patch had changed its signature, but since this is an EAPI,
we can't break it. Now ecore_fb_input_device_window_set() was added to
provide the same functionality.
SVN revision: 65589
Subject: [E-devel] [Patch][ecore_file] Fix buffer flush bug
fclose() should be called before callback is called to notify user of
download completion because it cannot be flushed sometimes.
SVN revision: 65572
ecore_evas_fb.c: In function ‘_ecore_evas_fb_init’:
ecore_evas_fb.c:248:14: warning: implicit declaration of function ‘ecore_fb_ts_event_window_set’
Signed-off-by: Mike McCormack <mj.mccormack@samsung.com>
SVN revision: 65546
AFAIU, this should be the correct, although this parameter isn't used
anywhere. But let's keep some consistency with the other backends, and
try to avoid future bugs.
Patch for SiT.
SVN revision: 65540
mouse is in on init (as events wont always give this) and focus
is set on show if appropriate if no focus in/out events come
from the back-end later
Fix setting override state to only hide if it should be
visible at that point in x back end support
SVN revision: 65508
Shift (left/right), Control and Alt are now detected.
AltGr is not, as it's actually a Control (left) key,
immediately followed by an Alt key, and I don't know how
to see know if the left Control kye is the one pressed by
the user or the one thrown by AltGr.
modifiers are not set too. Will be set soon
SVN revision: 65415
code which were not setting icccm wm protocols correctly. This 'fix'
matches up w/ the xcb_icccm_* functions defined in the xcb headers.
SVN revision: 65195
Fix warning: large integer implicitly truncated to unsigned type.
In ecore_x_window_prop_card32_list_get(): return -1 if malloc fails.
Patch from OpenBSD via Jonathan Armani
SVN revision: 65075
xcb_get_property_unchecked() requires an uint32_t as last parameter, so use
UINT_MAX instead of LONG_MAX
Patch from OpenBSD via Jonathan Armani
SVN revision: 65073
Subject: [E-devel] [ecore_con] Bug fix when data is NULL.
If data is NULL, curl makes length field as -1 .
length filed should be set by 0.
SVN revision: 65006
Subject: [E-devel] [ecore_file] Check a return value of
ecore_con_url_get()
_ecore_file_download_curl() does not check return value of even if
ecore_con_url_get(), it can be failed.
I add a simple error checking.
Please review this patch.
SVN revision: 64795
Subject: Re: review request : ecore x patch for X Gesture extention
Do you remember that I told you X gesture extension patch for ecore x ?
I’d like to put the attached patch to ecore_x in upstream.
This patch is just for initializing/receiving X gesture extension stuff.
Would you please put this in SVN ? : )
Thanks and regards,
Sung-Jin Park
SVN revision: 64732
Subject: Re: [E-devel] [PATCH] Fix the bug that user cannot input key
At this moment, space is processed in case of WM_KEYDOWN and WM_CHAR.
Therefore, 'space' key is processed twice, it makes bug.
In previous my patch, I changed keyname and string wrong.
I send the revised patch again.
(also applied to wince)
this is am improvement.. but its not all "right" yet. :(
SVN revision: 64597
Subject: [E-devel] Small patch with ecore_con_url.
When I tested with elementary map, there is too many logs on console.
It makes me difficult to distinguish normal logs that I want to debug.
:'-(
It's libcurl's internal log for representing progress bar like this.
-----
downloaded : 343 / 363
% Total % Received % Xferd Average Speed Time Time Time
Current
Dload Upload Total Spent
Left
Speed
0 0 0 0 0 0 0 0 --:--:-- --:--:--
--:--:--
0
downloaded : 343 / 362
% Total % Received % Xferd Average Speed Time Time Time
Current
Dload Upload Total Spent
Left
Speed
0 0 0 0 0 0 0 0 --:--:-- --:--:--
--:--:--
0
downloaded : 343 / 361
% Total % Received % Xferd Average Speed Time Time Time
Current
Dload Upload Total Spent
Left
Speed
0 0 0 0 0 0 0 0 --:--:-- --:--:--
--:--:--
0
downloaded : 344 / 365
downloaded : 345 / 365
downloaded : 346 / 365
downloaded : 347 / 365
-----
So, I want to remove this log.
And I look around ecore_con_url's code.
Then I found solution for it.
SVN revision: 64596
The reson why I add this is for communicate with X in async mode.
For example, If applications call elm_win_rotation_with_resize_set API
when they start run and rotation mode is set.
ecore size doesn't changed yet, so it make elm window size to 1
becaues elm window's resize function use ecore_evas_geometry_get API.
so I add new api help upperside get info related with recently request geometry
SVN revision: 64492
Nothing changes, only making the ecore fb engine to send keyboard and
mouse events using ecore_input_evas, instead of its own ecore events.
Patch for SiT.
SVN revision: 64447
There are two main reasons for disabling this now:
- it's not working, some problem with an assertion of ecore_lock is
making the program abort;
- it's wrong, mouse move events can only be filtered out if they are
received from an absolute axis device (touchscreen and so), not
relative ones, like mouse or touchpads.
Patch for SiT.
SVN revision: 64444
Jihoon Kim
윤정현
Rafael Fonseca
Christopher Michael
Mike Blumenkrantz
Tom Hacohen
Carsten Haitzler
Daniel Juyung Seo
Cedric BAIL
Gustavo Sverzut Barbieri
Youness Alaoui
Guillaume Friloux
Jiyoun Park
Leif Middelschulte
Stefan Schmidt
ChunEon Park
Sanjeev BA
Bluezery
Vincent Torri
Iván Briano
Doyoun Kang
Mike McCormack
Mike McCormack
Rafael Antognolli
Sung-Jin Park
Nicolas Aguirre
Jihoon Kim
Kim Yunhan