Bug #1456627 “Almost unusuable while non-Ubuntu device connected...” : Bugs : qtcreator-plugin-ubuntu package : Ubuntu

Revision history for this message

Florian W. (florian-will) wrote on 2015-05-19:

#1

While this happened all the time in the last few days, I now did something that apparently made it stop (temporarily?).

Maybe deleting the device from the Devices view in Ubuntu SDK? It re-appeared quickly, but maybe that stopped it. Maybe it's important that the same device (Nexus 4) has been used with the Ubuntu OS instead of Android before, using the Dual-Boot mechanism.

Revision history for this message

Zoltan Balogh (bzoltan) wrote on 2015-05-19:

#2

First of all, thanks for this report.

I try to corner all the issues you mention, butr first let's sort out few things:

1) "qtcreator started using a complete CPU core soon after startup"

What CPU and how much memory do you have on that device? Were you monitoring during that "hanging" period what else was going on in your system? Anyhow, QtCreator is a big and heavy application, it is not free from bugs. If we find a resource leak, we will follow it up with the upstream developer. Please note that QtCreator is downstreamed product of the QtCompany

2) about the device_version
If you check the /usr/share/qtcreator/ubuntu/scripts/device_version script it does very little. It barely executes more than 6 simple commands to figure out about the attached device. It does not use much CPU or any system resources. In your case it got confused because your android device had the same ID as the earlier registered Ubuntu device (due to the dual boot). Anyhow the SDK tries three times to discover the attached device and then it gives up.

3) About OnlyFindGrepItself
Sorry my ignorance, but I am not a kernel guru. How does that script proves that the qtcreator is forking 30 times a second? I would ps the qtcreator process and follow if the PID is changing. The fact that the grep gets higher PID number at each run is the business of the kernel. I doubt that it indicates any problem with the QtCreator. But I would be happy to learn about this case more if you would give some details.

One thing is for sure, that the SDK gets confused with that dual bootable Ubuntu/Android device and we will invstigate this use case and push out a fix very soon. But we need mpore information about the CPU load issue as with the given information I can only see it as a momentary stability issue of a tool what our SDK uses. It is very unlikely that this problem was caused by the qtcreator-plugin-ubuntu plugin.

Changed in qtcreator-plugin-ubuntu (Ubuntu):
status:	New → Incomplete
importance:	Undecided → Medium
assignee:	nobody → Zoltan Balogh (bzoltan)

Revision history for this message

Florian W. (florian-will) wrote on 2015-05-19:

#3

Download full text (4.5 KiB)

Thanks for your important questions, Zoltan!

> What CPU and how much memory do you have on that device?

It's an old AMD Phenom II X4 955 BE. Still, qt creator usually needs much less than one full core when just editing code files and generally works quite well. I have 8GB of memory.

> Were you monitoring during that "hanging" period what else was going on in your system?

I looked at "top" and found that qtcreator uses 99% (out of 400% theoretically possible with my 4 core CPU).

No monitoring other than that, but I reproduced the issue again. There are two different /usr/share/qtcreator/ubuntu/scripts/device_version processes running right now, and in total I counted more than 3 of them since I started ubuntu-sdk. At least that's what I assume, I've repeatedly called ps aux | grep script and each time, there are a few of them active and they have new PIDs each time.

> Anyhow, QtCreator is a big and heavy application, it is not free from bugs. If we find a resource leak, we will follow it up with the upstream developer.

It's working quite nicely when I don't have my Android phone attached via USB, so I don't think it's an upstream bug. Also worked fine when I attached my phone with Ubuntu booted on it.

> [The device_version script] does not use much CPU or any system resources. In your case it got confused because your android device had the same ID as the earlier registered Ubuntu device (due to the dual boot). Anyhow the SDK tries three times to discover the attached device and then it gives up.

Yeah, the script itself seems to be relatively lightweight. However it's executed in what appears to be an infinite loop on my system. So the "three attempts" limit might not work correctly?

> Sorry my ignorance, but I am not a kernel guru. How does that script proves that the qtcreator is forking 30 times a second?

Sorry, my post was not even close to clear enough about that. AFAIK the shell command shows that there were ~300 new processes spawned within the 10 seconds sleep period, because the PID of each grep call was incremented by ~300 compared to the last grep. AFAIK, PIDs are chosen incrementally, so each new process gets a PID of $lastPID+1, unless that PID is already in use. So a "grep, wait 10 seconds, grep" sequence should show that both grep PIDs are pretty similar on my relatively calm desktop system. A difference of 300 between both PIDs tells me that my system spawned 300 processes in 10 seconds. Since top shows nothing unusual other than 99% qtcreator CPU usage, I assumed that most of the 300 processes were spawned by qtcreator (I suspect: by the qtcreator ubuntu plugin).

I have now disconnected my phone from USB. top shows that qtcreator is now down to 0.0%, but takes up 30% of my memory (ps shows RSS 2468524 – 2.4G, this might indicate a leak? Can't remember what qtcreator memory usage is normal after 30 mins of idling in the background.).

Now let me repeat that command:
$ while true; do ps aux | grep OnlyFindGrepItself; sleep 10; done
x 6171 0.0 0.0 18980 2136 pts/0 S+ 21:23 0:00 grep --color=auto OnlyFindGrepItself
x 6174 0.0 0.0 18980 2192 pts/0 S+ 21:23 0:00 grep --color=auto OnlyF...