use all memory then crashes once first midi message is sent #43
Comments
|
Hi Can you run it in debug mode and paste some of the last few lines of output? Thanks Dave |
|
ok, Ive just realised its actually happening even if I don't connect it to the mac.. Im trying to get it to crash now in debug... currently it seems to be filling up the log file with the same message about poly aftertouch (note: Ive stopped pressing anything at this point) hmm, I guess what I need to do is test with another controller , perhaps its got an issue with polyAT? |
|
ok, it's something about that controller. I was testing using an Ableton Push 2. Ive now switched to another controller and it seems to be working fine! |
|
ok, this is odd... so the new control (Access Virus ) works fine for a while, and then suddenly I get a stuck note (so note_on , but no note_off) now the odd thing is... this controller does not send poly AT messages! |
|
those 'aa' bytes look suspicious ;) |
|
The first 0xaa byte is a fake one that I prepend to the input buffer to allow me to use the same MIDI parsing that I have for the raw local port. From what you're showing me, it's highly-likely that I'm prepending that to every read from the ALSA buffer. I'll dig into that. |
…s the internal origin prefix to be included more than once
|
Can you check out the latest experimental branch and confirm if the fix I added solves the problem? Thanks Dave |
|
k, tried with experimental include last fix, unfortunately same issue. so what I see is... one thing, that may be relevant... (and I could see now it stayed up) the controller is sending 'active sense' messages constantly... perhaps this is connected? BTW: is there a way to for raveloxmidi to read from virtual alsa devices... or for me to send messages to it from a process running on the rPI (e.g. connect via aconnect) |
|
Can you provide me with the full debug output somewhere so that I can take a look? Regarding using aconnect, If you take a look at FAQ.md, I wrote up some brief instructions on testing input support using aplaymidi. Thanks Dave |
|
I should have added that I don't do anything with active sensing messages. I receive them but that's all. |
|
doh, yeah of course snd-virmidi forgot all about that :) active sensing - cool, i just wanted to point out they were there, as its constant traffic. ok, here is the debug... so as mentioned previously, the debug mode slows it down, so I don't actually wait for it to run out of memory... but instead i just wait till i can see it steadily growing. so in this particular case what i did was: then I killed it... and zipped up the log file. Note: I did not even bother to connect it to the mac. so the messages were going nowhere. the reason i dont connect is it doesnt affect if it crashes or not |
|
I may have had some logic reversed in the change I made. Can you check out the latest experimental code and send me the logging again? I've added a log line that should contain the text "first_byte_ignore", if you want to grep for that. Thanks DAve |
|
I've fixed up my logic. Could you run another test please? Thanks Dave |
|
k, pulled, compiled, tested - issue remains |
|
I had a quick look at the logs, so issue would appear to occur around line 5900 what we can see is we get a new note_on (90, 3a, 19) on the next active sensing , we can see in the hex dump that 3a,19 remained (with a new active sensing message also present) raveloxmidi seems to then always leave these present in the buffer, and tries to continue to process active sensing messages, but by looking at previous processing msgs, it appears this is incomplete. relevant portion of log |
|
Could you check out latest experimental and test again. It looks like I'm not advancing forward after reading MIDI commands that have no parameters. Thanks Dave |
|
done - now it just segfaults as soon as it starts (both with and without debug) |
current_len looks very dodgy ;) |
|
the issue is caused by your latest change, using current_len--, |
|
Yep, me culpa :) Please try again with the latest experimental code. Thanks for your patience. Dave |
|
this looks good ....... no issues with this device now :) ok, Ive got another issue, when lots of messages are sent but I'll raise that separately as its quite likely not related?! |
|
We'll call this one closed and move on to #44 |
Hi,
im attempting to run raveloxmidi on a rPI3 connecting to a mac (10.14.5) over wifi
a very simple setup.. built from GitHub
sudo raveloxmidi -N -d -c /etc/raveloxmidi.conf
it starts ok on the rPI, reporting
and the mac sees it, and then I can connect...
viewing with 'top' on rPI, all looks good,
but as soon as i touch a key on the midi controller
the cpu % rises to 100%, and memory then rises to 100% and ravelosmidi is killed by out of memory reaper, and I can see below in dmesg
thoughts?
The text was updated successfully, but these errors were encountered: