so RS ended up with 200+ zombie child processes. This prevented gpg from being called.
The fix was to add waitpid() calls to cleanup the zombies.
* added cleanupZombies() to authgpg.cc (OSX only).
* updated OSX gpgme libraries to 1.3.0 (added libassuan to RetroShare.pro).
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3974 b45a01b8-16f6-495d-af2f-9b41ad6348cc
- authssl would call OPENSSL_malloc without OPENSSL_free
- pqistore was not deleting items in pqiSSLStore when BIN_FLAGS_NO_DELETE is not here
- rsdir/rsinit were calling opendir withoug closedir (which might eat FDs, in addition)
- udplayer was not freeing it's 16KB buffer at end of loop.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3948 b45a01b8-16f6-495d-af2f-9b41ad6348cc
Changed the usage of RsPeerDetails.validLvl to RsPeerDetails.trustLvl for the display in the column "Did I authenticated peer" in NetworkDialog.
Added an icon to the MessageBox of the NotifySysMessage.
Fixed german language.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3926 b45a01b8-16f6-495d-af2f-9b41ad6348cc
* shifted dht stats call outside of p3connmgr's mtx.
NB: As Dht callback functions hold DHT lock, then lock p3ConnectMgr...
we cannot have any opposite locks (p3connmgr, then DHT) or we have
potential for a deadlock.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3861 b45a01b8-16f6-495d-af2f-9b41ad6348cc
* new uniform #define DEBUG_TRANSFERS to switch on packet timings.
* switched off by default.
* enabled BinData printing again.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3785 b45a01b8-16f6-495d-af2f-9b41ad6348cc
* print packet size and time - when actually sending it.
* print reason for stopping to read or send packets.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3780 b45a01b8-16f6-495d-af2f-9b41ad6348cc
Reload all gpg keys in AuthGPGimpl::LoadCertificateFromString, when new signatures are added - not only the key was imported.
Changed parameters of RsPeers and AuthGPG from "std::string" to "const std::string&"
Recompile of the GUI needed.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3753 b45a01b8-16f6-495d-af2f-9b41ad6348cc
* Plumbed stats through the the retroshare interface.
* Also changed Channel store period back to 10 minutes.... (Reduce overall file count).
This must be done some time before the any changes are made to Channels periods.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3721 b45a01b8-16f6-495d-af2f-9b41ad6348cc
- FAIL if too old.
- FAIL if invalid address.
- FAIL if on BannedIpList (=1.0.0.0)
- ACCEPT if extAddress.
- ACCEPT if !extAddress && sameNet(localAddr).
- FAIL all the rest.
* Test Results are printed if CONN_DEBUG is switched on.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3705 b45a01b8-16f6-495d-af2f-9b41ad6348cc
Problem:
RetroShare connects to all ssl's at once. In Windows XP there is a build-in connection limit of concurrent incomplete connections.
Quote:
"The TCP/IP stack in Windows XP with Service Pack 2 (SP2) installed limits the number of concurrent, incomplete outbound TCP connection attempts. When the limit is reached, subsequent connection attempts are put in a queue and resolved at a fixed rate so that there are only a limited number of connections in the incomplete state."
This results in a high usage of the nonepaged pool and when it runs over the limit it results in the following error in the system log:
German: "TCP/IP hat das Sicherheitslimit erreicht, das für die Anzahl gleichzeitiger TCP-Verbindungsversuche festgelegt wurde."
English: "TCP/IP has reached the security limit imposed on the number of concurrent (incomplete) TCP connect attempts."
After a longer runtime of RetroShare and a very high usage of the nonepaged pool all connections of all running programs are lost and no new connections can be made. At the end it can cause a blue screen.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3631 b45a01b8-16f6-495d-af2f-9b41ad6348cc
* NOTE about how to correctly implement Peer filtering.
* added peerConnectRequest() when PeerCallback(success) happens... starts UDP connection.
* Split the p3conmgr::tryConnect() functions into sub functions.
* added p3ConnMgr functions for UDP connections.
* Added AddifUnique(conect_address), to ConnMgr, this should reduce the number of connection attempts, by removing dups.
* Disabled bitdht in libretroshare.pro for the moment.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3605 b45a01b8-16f6-495d-af2f-9b41ad6348cc
Added second mutex for the gpg engine.
Now the call to the gpg engine (for example AuthGPGimpl::LoadCertificateFromString) doesn't block the GUI.
The existing problem is still the call of AuthGPGimpl::storeAllKeys every minute before access the gpg data.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3555 b45a01b8-16f6-495d-af2f-9b41ad6348cc
- added "mStatusChanged = true" on disconnect of a peer - the statusChange was not notified in time
MessengerWindow
- removed unused method updatePeersAvatar
Changes in status service:
- removed unused method statusAvailable
- new notifier for status change
- enabled the tick for receiving items
- always up to date status map
- no huge memory usage, if no list is visible who call getStatus and processed the received items
- send notify on status change
- renamed getStatus to getStatusList and created a new method getStatus to get status of one peer
- fixed: send status directly after the connect of a peer
Now the frame around the avatar in private chat changes when the peer change the state.
git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@3380 b45a01b8-16f6-495d-af2f-9b41ad6348cc
