4b873fa015
PVAClientRegister: automatically register "pva" provider
2017-06-01 01:57:30 +02:00
48b3b648ef
add ServerContext::getCurrentConfig()
2017-06-01 01:56:46 +02:00
65cdff288f
Configuration list keys
2017-06-01 01:54:17 +02:00
8971d2771c
ServerContextImpl::destroyAllTransports()
...
Better check for ref. leaks
2017-05-31 15:58:15 +02:00
3fa6a4e4cc
TransportRegistry composed
...
no need for a seperate allocation and shared_ptr
2017-05-31 15:48:30 +02:00
3ad27665e8
BlockingTCPTransportCodec::close() wait for threads
2017-05-31 15:47:46 +02:00
6ad8ba3ba2
AbstractCodec pure virtual redundant to Transport::close()=0
2017-05-31 13:09:21 +02:00
e72396a022
auto-start PVA server w/ initHook
...
split of this code into a new library pvAccessIOC
which depends on all core libraries.
2017-05-31 13:01:52 +02:00
634e50e011
startPVAServer() has to keep the pointer
2017-05-31 12:18:42 +02:00
ce25f0b175
ServerContext ref. loop breaking and threading
...
Drop unnecessary "worker" thread which does no work.
Ensure that returned shared_ptr is unique()==true.
Add ServerContext::create() to start a new server
with specific config and/or providers
2017-05-31 11:40:51 +02:00
0fbbcc2d9f
remove deprecated Properties
2017-05-30 17:33:04 +02:00
dd78b467f8
consolidate AtomicBoolean
2017-05-30 17:32:28 +02:00
277e5c33fb
privatize some API details
...
hide a number of APIs which should not
be public APIs.
2017-05-30 17:12:31 +02:00
7c85e6073b
move ServerContextImpl into private header
2017-05-30 12:07:23 +02:00
7c2a093338
add ChannelBaseRequester::channelDisconnect()
...
per-operation notification of channel disconnect/destroy.
2017-05-25 18:01:41 -04:00
d7a7b81900
move monitor.h from pvDataCPP
2017-05-24 17:20:17 -04:00
e80ab0464e
less hacky-ness in operation disconnect/destroy notification
...
avoid pointer/reference comparison.
2017-05-24 17:20:17 -04:00
24c9b31cf6
update pvAccess.h doc comments
...
describe how things should be!
2017-05-24 17:20:17 -04:00
3283bed413
drop PVA_ALIGNMENT constant
...
This will forevermore be 1, so no
reason to incur complexity of testing this.
size % PVA_ALIGNMENT -> 0
alignBuffer(PVA_ALIGNMENT) -> no-op
alignedValue(val, PVA_ALIGNMENT) -> val
2017-05-24 17:20:17 -04:00
87dca19708
codec: avoid indirection when accessing buffers
...
avoid some indirection to make this code easier to follow.
move buffer lower limit to base class.
2017-05-24 17:20:17 -04:00
0c1f410e63
avoid unnecessary auto_ptr/new
2017-05-24 17:20:17 -04:00
b05e7699b5
remove no-op lock/unlock from Lock sub-classes
2017-05-24 17:20:17 -04:00
bbe6cb2553
Lock provide default no-op implementation for lock/unlock
2017-05-24 17:20:17 -04:00
2cc1be2dee
codec.h: minor
2017-05-24 17:20:17 -04:00
ab4f0b7e3a
collapse BlockingSocketAbstractCodec -> BlockingTCPTransportCodec
2017-05-24 17:20:17 -04:00
e47124aa30
collapse BlockingAbstractCodec into BlockingSocketAbstractCodec
2017-05-24 17:20:17 -04:00
5428b1ed2d
codec.h: add virtual and decorate w/ OVERRIDE/FINAL
...
Mark where inheritance is actually happening,
and where it ends.
2017-05-24 17:20:17 -04:00
d404769133
add default for first Channel::createChannel()
...
default for first form which calls the second
2017-05-24 17:20:17 -04:00
568054ce68
add virtual dtors
...
being pedantic...
2017-05-24 17:20:17 -04:00
65cb1c8397
minor
2017-05-10 14:59:16 -04:00
9dbd354bc0
caProvider: pretend to allow create with Configuration
2017-05-09 18:23:45 -04:00
065d653a14
minor
2017-05-09 18:23:45 -04:00
9253c31c0b
rework ChannelProviderRegistry
...
Replace interface+impls with a single concrete container
which wraps a map.
This breaks code which was sub-classing.
However, no external code can be found which does this.
A build() method is used instead of simple ctor so
that the ctor is private, which will cause theoretical
external sub-class code to fail to compile.
Code using getChannelProviderRegistry()
and existing methods will behave as before.
Add ChannelProviderRegistry::getFactory() for pass through
to ChannelProviderFactory (Registery is just a proxy()...)
Remove existing get/create virtual method in favor of getFactory.
Update both existing implementations.
Add SimpleChannelProviderFactory to cut down on boilerplace
for the common case.
Add method to clear mapping and release any saved sharedInstance.
2017-04-25 21:15:43 -04:00
4a9d79f019
sockAddrToA -> sockAddrToDottedIP
...
Avoid blocking host name lookup for debug log messages.
Can cause long delays when DNS not available.
2017-04-15 11:48:24 -04:00
2cf38c875b
updated version
2016-07-26 23:34:15 +02:00
c12b0d130e
Merge pull request #57 from dhickin/caProvider_arrays
...
caProvider: Remove control/valueAlarm for arrays
2016-07-19 17:26:48 +02:00
e0e45c3a0e
caProvider: Remove control/valueAlarm for arrays
...
This brings caProvider behaviour in line with pvaSrv.
2016-07-18 16:35:43 +01:00
dc61b3a4cd
Correctly set status in caProvider
2016-07-18 15:26:17 +01:00
3a73cb687b
ca provider destruction fixed
...
resolves #53
2016-06-14 22:34:14 +02:00
56286872b1
Correct get of display.limitLow
2016-05-18 11:38:23 +01:00
5d992aee1a
Keep NT id if optional field of DBR match not requested
2016-05-17 10:34:31 +01:00
362f3c02d7
Return requested fields in caProvider
...
caProvider matches to closest DBR type. Strip out unrequested fields
and only return requested parts of structure fields. Includes enums.
Check fields are non-null rather than throwing. Check labels non-null to
handle requests for choices field without index.
Fixes #42 . Covers most cases of #41 .
2016-05-17 10:33:24 +01:00
4e264f2295
win32: moved osiSockAttach() call before loadConfiguration
...
should fix #45
2016-04-26 20:48:27 +02:00
a9e3f274c0
remove dead code
2016-04-19 15:10:41 -04:00
998bf8ac67
PVAServerRegister:: allow providerList, do not use epicsAtExit
2016-04-19 10:07:58 -04:00
0f61408a28
Minor clean-up after issue #43 fix.
2016-04-14 11:36:14 -05:00
6dc9b0f96f
replaced LGPL wildcard matcher w/ simplistic epics-base version
2016-04-14 11:31:32 +02:00
6698bb86fb
Additional includes needed for MSVC
2016-04-06 17:02:03 -05:00
b928ea4aab
Fix oops when MB turned off.
2016-04-06 16:21:06 -05:00
9a8644033d
Prevent redefinition of NOMINMAX
...
While this protection is not strictly necessary here, if someone
does a cut-and-paste into a header file it should be included.
2016-04-06 15:22:11 -05:00
Michael Davidsaver