atlesn / rrr Goto Github PK

• SNMP manager module
• SNMP agent module

The test suite gave the following error (has happened once):
test: ../nptl/pthread_mutex_lock.c:81: __pthread_mutex_lock: Assertion `mutex->__data.__owner == 0' failed.

Some mutex-stuff is badly implemented, possibly a mutex is being unlocked by another thread than the locker.

Server should support HTTP->WebSocket upgrades with persistent connections.

• Basic negotiation handled by httpserver
• Higher level messages sent out of httpserver for processing in other modules like Perl
• Higher level replies from other modules

libtool: error: error: cannot install 'mysql.la' to a directory not ending in /usr/local/lib/rrr

Currently most modules does not properly destroy their FIFO buffers because the mutex is used by multiple threads (it's the mutex which is problematic to destroy). Memory isn't leaked, but pthread_mutex_destroy is not called. Figure out how to solve this.

Title says it all, look for TODO inside p_influxdb.c.

Make user able to choose using URI to specify server, username etc. in the configuration file.

Working with RRR arrays in Perl i cumbersome, needs improvement/abstraction.

There are some warnings involving function pointer casts with (PyCFunction), fix.

python branch. Python module is defunct with new fork system. Code must be re-organized.

• All python3 code must reside within the process/source forks, nothing in main thread
• Reduce python3 socket class to a wrapper which puts messages into mmap channel

Figure out what features of an HTTP server or client which can be useful.

It might be useful to integrate RRR arrays and send/receive POST and GET data.

systemd branch. Some structure in the build system is already made but is not complete. Files in misc/systemd . RRR must be able to start many config files from one process. This now works, but there are some issues with thread shutdown.

Not currently done, see mqtt_assemble.c. Also need to figure out what's reasonable to send.

The FIFO buffer currently goes through some consistency checking to detect errors. This is done every call and is slowing things down. The buffer needs to be stress tested in some way and the consistency checks disabled, although there hasn't been any problems with it in a while.

This is both in preparation for later manuals/guides and to test the program and find bugs and things which need to be changed.

Currently all session states are stored in the RAM engine. The session system is already designed to have multiple different storage engines. A database engine, e.g. MySQL, will allow for client state to be preserved across restarts.

Packets like UNSUBSCRIBE is currently not handled because the code isn't written. Finish this off. Stuff like retain also isn't written.

Use the code coverage scripts to detect code not included in the test suite, and extend the tests. Possibly have more "message channels" in the test to avoid complicating what's already there even further.

The manual specifies that the default value of the parameter mqtt_publish_rrr_message is yes, but in the code it is by default set to no.

Make duplicator automatically convert between ip_buffer-messages and rrr_message, allowing senders og readers to use either.

Nov 11 00:57:09 label-print-01 rrr[13636]: Error from shmget in __rrr_mmap_channel_allocate: No space left on device
Nov 11 00:57:09 label-print-01 rrr[13636]: Could not allocate memory in rrr_mmap_channel_write
Nov 11 00:57:09 label-print-01 rrr[13636]: Could not send address message on mmap channel in __rrr_cmodule_send_message name
Nov 11 00:57:09 label-print-01 rrr[13636]: Error while sending message in rrr_cmodule_send_to_fork

------ Shared Memory Limits --------
max number of segments = 4096
max seg size (kbytes) = 18014398509465599
max total shared memory (kbytes) = 18014398509481980
min seg size (bytes) = 1

All modules restarted correctly afterwards.

The function rrr_mqtt_topic_match_str does not work correctly, it might ignore the last character of input data.