Age | Commit message (Collapse) | Author | Files | Lines |
|
Change-Id: I056b5953cff53c6053412b56c6be4a31aa7c33ba
|
|
Change-Id: I96e0793552eb298334ec6483a6e4c3a1f0d12cc1
|
|
[model] Redwood,Kiran,B3(Wearable)
[binary_type] AP
[customer] Docomo/Orange/ATT/Open
[issue#] N/A
[problem]
[cause]
[solution]
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I46bb8c0de37684f42ce59511816a7149cc7ab9b5
|
|
into tizen
Conflicts:
CMakeLists.txt
include/com-core.h
include/com-core_internal.h
include/com-core_packet-router.h
include/com-core_packet.h
include/com-core_thread.h
include/packet.h
include/secure_socket.h
packaging/libcom-core.changes
packaging/libcom-core.spec
src/com-core.c
src/com-core_packet.c
src/com-core_thread.c
src/packet.c
src/secure_socket.c
Sync with the latest private implementations
Change-Id: I8aec4d532b687226d29eb11c18332a2cabc4f5e3
|
|
between local processes.
[model] Redwood,Kiran,B3(Wearable)
[binary_type] AP
[customer] Docomo/Orange/ATT/Open
[issue#] N/A
[problem]
[cause]
[solution]
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: Ifbcbae54f95c176bf585d3be2f6266460bd27434
|
|
If a command string is started with 0x01,
it will be treated as an integer type.
It could be used as an index of the command table.
It will increase the command handler searching time.
[model] Redwood,Kiran,B3(Wearable)
[binary_type] AP
[customer] Docomo/Orange/ATT/Open
[issue#] N/A
[problem]
[cause]
[solution]
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I6a6be27e01ac0e00b2124ac1132d6b6e466bf99d
|
|
Change-Id: I58823aa462ac6728773095448214fac58652214a
|
|
Change-Id: I50c9f6f02580d8390307045278fdef0999a65cbb
|
|
From the recv ctx callback, the socket can be closed.
But the glib cannot detect the closed FD correctly.
(It doesn't toggles G_IO_HUP, G_IO_ERR, G_IO_NVAL bits even though I closed the FD)
So I checked the FD from the callback explicitly, before return from it.
However this patch will be available only for the Single-Thread loop. (com-core)
And the receive context is optimized.
Every single loop after done to prepare A receive context,
it will be destroyed. But this patch will reuse it instead of destroying it.
So there is no need to reallocate same memory again.
It will save few milli (or micro) seconds.
for the future-work.
I'll optimize the "packet" creating code.
It can be reused too after finish the processing of receive context.
Then we can save more few milli(micro) seconds (for reallocating time).
Each execution takes small time.
But this scenario occupies almost time for communicating with other processes.
it means, even though it can save few milli(micro) secs for each time.
The total saved time will be meaningful.
Change-Id: Ibb3adc715090b6d78a5cfeb48d28463f35b9a0f4
|
|
dlist_remove_data always find the data before delete the item.
This patch will find the item first like dlist_remove_data.
but if there is no item in the list,
do not release it. just skip the destroy function call.
Change-Id: Ia15fa5c7dc3f283a19e7b5109c3c5796623b55d2
|
|
for the ugly GCC 4.6
Change-Id: I39d6d6666360f2c66896b14ce7e95c71b15c2496
|
|
Change-Id: I29dc6393c1891d74efbd886ccf36dcb58eff42e0
|
|
Change-Id: I9bcb9fa619807d972bda86d795e0e984c1091578
|
|
Sync with tizen 3.0 - supporting systemd socket activation
Change-Id: I8fdbb2da60273a76c2493230eb41140105e6438b
|
|
Disable O_NONBLOCK option for creating a "pipe".
It makes unpredictable busy waiting problem when it fails to write data to pipe.
In some cases, the consumer thread could not empty the pipe,
the producer thread failed to write with EAGAIN error if the pipe created with O_NONBLOCK.
then the producer thread fall into the busy waiting status.
This patch will keep the producer from busy waiting problem.
Change-Id: I34d1433b4c380deb2bc8f03d476943652de6c23f
|
|
Change-Id: Ic171862e4f09609d6a69c5c067907638d7ac40a6
|
|
Change-Id: If94ce60148eab509fc21f5696d046528a35a9aba
|
|
Change-Id: I48fea5c3dc6d146cc6118f77c61d89569d6058ea
Signed-off-by: Patrick McCarty <patrick.mccarty@linux.intel.com>
|
|
Change-Id: I5b6584acfce760139b8ffada4e469d150ec2723e
|
|
Change-Id: I3a934684888ae00e66b9d374940ef7eb267a3582
|
|
Restore use of %cmake in the spec file so that variables like
LIB_INSTALL_DIR are passed to cmake.
Change-Id: I3ed0b40417728ccbf4e171492ec9a7fc77743097
Signed-off-by: Rusty Lynch <rusty.lynch@intel.com>
|
|
ssh://review.tizendev.org:29418/framework/appfw/com-core into work
Applying socket activation should be supported by new API or Schema.
The latest com-core supports the inet & ipc socket communication using schema.
To open/connect a socket, user can select inet or ipc using schema.
remote://IPADDR:port
local://SOCKET_FILE_PATH
practical example
remote://192.168.0.1:1234
local:///tmp/.my.socket
So I just discards the systemd socket activation code now.
I think, it should be supported by new socket URI.
such as
systemd://SOCKET_FILE
This new schema should be added to handle the socket activation mode.
Conflicts:
src/secure_socket.c
Change-Id: Ib2c0d1c7f2d6b0337cd408cba11164626ca87ff4
|
|
Change-Id: Ie1e70c0576c9381833981080855cfb9972ae582d
|
|
Change-Id: Ice0e00e2ffbba6f5a31012b56cf6df0f4884db95
|
|
Change-Id: I14c4d05d37c792834ce73aa6a3df2c972a488650
|
|
Change-Id: I306d315b59a56feb2474bd9a460a50c821aaa50e
Signed-off-by: Rusty Lynch <rusty.lynch@intel.com>
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem]
[cause]
[solution]
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I8b8cc17a189e72b0009ec40fbd3ef3adbe13314e
|
|
Even if a block has only one line, it should be begin with { and end with }
Change-Id: I99c7f6d43cd0b673d2d912dd1835d2b57ad1de66
|
|
Use the '{' '}' for local block(if, for, while, ...) even if it has A line.
Change-Id: I29a8d8ab2c6023db66312b2d2ca93b8c4440248a
|
|
Change-Id: Ic3701c4181f61016bd6375844087ecf9480ce746
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem]
[cause]
[solution]
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I67f778a2006342b055290bc67c38a88d4ac7d6d2
|
|
Change-Id: I959c74dc3c3ddc442176d22ff0df5a3d687b983d
|
|
Recv bufsz: 512KB (kernel will double this to 1MB)
Send bufsz: 256KB (kernel will double this to 512KB)
This change is only can be applied by selection of kernel.
So even if I change this, the kernel can change it what it can support.
Change-Id: Ic3119dd82bbaf3845ac34fcf39f2f099cb9a70eb
|
|
crashed.
Because the disconnected callback is called and it will delete the receive context.
but the ack callback uses resources in the receive context.
This patch will prevent from crash.
Change-Id: I9015d04b6695e86c532b399de46bdc1fa6f31317
|
|
While in the recv callback, if the client tries to disconnect from the server,
the recv callback called again. because of disconnected event callback.
This patch will prevent call the same recv callback if it is in process.
Change-Id: Ib9a14849e0ff60ee72770d540bce31ac2c0cc396
|
|
Change-Id: I672c28932e275e6d28fe158c1c2c18ee18172cc1
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem] Some compiler (which keeps old standard) complains the last trailling comma.
[cause] It's a standard.
[solution] Remove the last trailling comma.
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: Id6f59d3caef17116c24a58a186f373c5e942ca01
|
|
Change-Id: I8bcae93851d6fff94c932363d12467ea7711092d
|
|
Change-Id: Ic830d5000575d9cd14012589e040ecf2248d70d8
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem] Reduce logs
[cause] Too many logs are printed.
[solution] Delete unnecessary logs.
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I84c4e3885d39d343bbd5f0ae11ee4e0b3ac9fa25
|
|
[model] Tizen
[binary_type] AP
[customer] Tizen Developer
[issue#] N/A
[problem] Hangs while terminating thread base com-core.
[cause] When the socket descriptor is used in a thread via "select", main thread should not close it.
[solution] Use the event-pipe to escape from select.
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: Ied7a8eccb4fc7ccb1ea180c1c79bed73d800087e
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem] Disconnected callback is not invoked.
[cause] When a main thread tries to close the connection, disconnected callback is not invoked.
[solution] Guarantees call of registered disconnected callback even though the main thread explictly close a socket.
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: Ia90480aca9e558d6edeb7b5a9656f89fc0fa5cc4
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem] Thread is not terminated. so the process hangs while waiting thread termination.
[cause] If a main thread close the fd, service thread can not detect it.
[solution] Add ctrl-pipe to terminate the service thread. (to wakeup it from select)
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I9d253eefa1b5c3206046cfff1f16ae89798a94fa
|
|
Crash occured while printing the integer value as string type.
This patch fixes it.
Change-Id: I064a4643a1586ab3f2f8a69c62e60324ce9331fa
|
|
[model] Redwood
[binary_type] AP
[customer] Docomo/Orange/Open
[issue#] N/A
[problem] Crash occured
[cause] print an integer as string type.
[solution] Fix it
[team] HomeTF
[request]
[horizontal_expansion]
Change-Id: I20c6fdc2b56841acb3bf32b5eecac804853474b6
|
|
Change-Id: I51b980eaba16fe84d24e1a9a8494a7ad2e5ddc9f
|
|
Patch 3/3
Do not update chunk->size using return of recv.
Update logging message
Patch 2/3
Update com-core.
Remove pthread_cancel.
Patch 1/3
Put complains of the valgind to sleep calm.
Change-Id: I75503d691b09b70a891e12ed5342cb634ca814b6
|
|
Change-Id: I3c1fadce91bc2fed987a7b4c2f146b12bedc31df
|
|
- Fix hardcoding path.
- Use %cmake to set default paths.
Change-Id: I484687e5906fe33fdb435cc14f1a83cd751c67f1
|
|
2012 -> 2013
Change-Id: Ieef4c7ae0dedc85d2ed8a23c133a881e53b8ffe6
|