btnx: Send keyboard and mouse combination events with mouse buttons

[daou]

Hi,

Just compiled btnx 0.4.7 and btnx config 0.4.7 on ubuntu hardy 64 bits.
When I try to launch btnx-config from terminal i get a segfault :/

Strace output of btnx-config ends like this:

open("/usr/share/pixmaps/btnx.png/btnx.png", O_RDONLY) = -1 ENOTDIR (Not a directory)
open("/usr/share/pixmaps/btnx.png/btnx.png", O_RDONLY) = -1 ENOTDIR (Not a directory)--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV (core dumped) +++
Process 24231 detached
open("/usr/share/pixmaps/btnx.png/btnx.png", O_RDONLY) = -1 ENOTDIR (Not a directory)
mmap(NULL, 135168, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f3ebfd74000
mremap(0x7f3ebfd74000, 135168, 266240, MREMAP_MAYMOVE) = 0x7f3eb8ba7000
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV (core dumped) +++
Process 24231 detached

any clue ?

thx

It seems that there is an error opening the btnx logo. Note the path "/usr/share/pixmaps/btnx.png/btnx.png". This is incorrect. You have just stumbled upon an at least 6 month old bug .

Interesting that no one else has had problems with it earlier. I found the bug and fixed it. It will be released with btnx-config v.0.4.8, which I will release as soon as I get the finished Russian translations from Launchpad. Alternatively, you can get the fix from the SVN repository.

Thanks for bringing this up.

[daou]

That's how btnx started in the first place, but the amount of effort involved in maintenance and development was horrendous. If you would like to create and host a config list, I'm sure it would be appreciated, but it would also be both rigid and redundant. A better choice for me was to list the individual key-bindings that people have been using successfully.
In fact, please feel free to add to my list if there are any that you are using which I haven't including.

Thanks,
blazoner

I updated the link on the first post to the new list you posted a few weeks ago.

And continuing your answer concerning jamesford's question:
It was very nice before with the ready config files. Two commands and the mouse was working. However, as blazoner mentioned, it ended up being a lot of (mistake prone) work for me to add the mice and their button codes.
Also, as I added more mice, I noticed that it was impossible to keep it fully automatic. Many mice use the same vendor and product IDs which is how btnx detects the mouse. So the user would still have to choose their mouse from a list (in which case they had to know the exact model of their mouse). Other problems also creeped in. For example, the MX Revolution mouse has at least two different revisions with different product IDs. It would end up being confusing to the user to choose from two different MX Revolution configurations.

But I'm considering bringing back the functionality in some form if I have enough time. I will talk about this in a later post that is a reply to blazoner's earlier comment (root button detection, user specific config files).

[daou]

...
Exceptions/Probelms:

1) The top button (behind the scroll wheel) is mapped by default to the Tracker Search tool in Gnome. So during the button detection it kept coming up. BTNX detected the button fine I just had to keep closing the search window that popped up. I haven't went back and remapped that button within Gnome.

2) The middle click from the top scroll wheel didn't get recognized during the button detection, but it seems to wrok fine afterwards...

Remapping the search button in Gnome should help with the Search tool popup. Also, the middle click probably started working after you used revoco: before revoco, the middle click only switches the mode. When changing to mode with revoco, you can get the middle click to act like a button click again and detect it with btnx (this is done in the mouse firmware).

[daou]

I've upgraded to hardy, and to the 0.4.7-1.deb packages of btnx and btnx-config, and I'm having the problem where X still seems to be intercepting my buttons;I use my thumb buttons for tab switching, but firefox is grabbing back/forward from them as well, which means I can't go right one tab without going back on the original tab. Very annoying.

The only suggestion I've seen so far is to go to 0.4.7, but as I said I've done that. Any other ideas?

Have you tried editing your xorg.conf file as suggested in the btnx manual troubleshooting section?

[daou]

I was thinking, (I know, that's dangerous..... )
Since USB mouses report their buttons, I don't see a reason for making button naming a root process. Rather, why not have a btnx-detect process that is run once as root, then have a user-level btnx-config that btnx reports the button presses to. This would avoid duplication of work for the process, and the end-user. And again, the end-user can do absolutely anything their access would allow.


In summary:
btnx-detect = a one time root process for initial detection of the mouse
btnx-root-daemon = a root process that listens to event handlers, and publishes them directly to the end-user application
btnx = much the same as now, but listening to btnx-root-daemon instead of the event handlers, thereby not needing to be a root process
btnx-config = much the same as now, but not a root process. Integrated into mouse properties, with command execution based upon UID

I know all this would require massive changes, but it would probably make the whole process more secure.

Feel free to pick the idea apart....

You've pretty much hit the nail on the head considering what I've been thinking about for later additions to btnx.

btnx-config would be a user application. The "Detect mouse + buttons" would launch what you call btnx-detect, a root process. This detects the mouse and buttons, and writes a configuration file to ie. /etc/btnx/mouse1_config. When a user opens btnx-config, it uses the root owned mouse config file as a skeleton, which the user fills in and saves to his home directory ie. /home/johndoe/.btnx/config1. These skeleton files can also be downloaded from elsewhere for mice that other people have detected. This does bring about its own problems, but it's an idea.

Your suggestion regarding btnx and the btnx-root-daemon is something I haven't thought of before, and I find it quite interesting. It would add a necessary layer of security to things like command execution as spoofed users (ie root). One thing it doesn't prevent, however, is keyloggers. I'll have to think about this more.

Thanks for the suggestions, and keep thinking (no matter how dangerous it might be ).

EDIT: There is a way around the keylogging problem, so perhaps that won't be an obstacle. But these are good ideas. I'm thinking about setting up an idea collaboration environment where we could discuss these things in a more organized way. I have a bad habit of forgetting what has been talked about in this thread because of its size. Maybe a wiki, or if Google has something useful... perhaps Launchpad blueprints. Let me know if you have ideas. As of now I keep everything in a text file in my btnx source directory. Current contents:

btnx:

• multiple mice configs
  • try different configs, do not die immediately if the default mouse is not detected
  • required for roaming users
• integrate btnx & btnx-config into a single package

btnx-config:

• test button for command execution to quickly test a command
• change UID to user name, make the command use su to change users
• allow multiuser, personal configurations
• white logo background
• refactor/rewrite in C++ (postponed):
  • implementing new features is becoming difficult and error-prone, no clear structure
  • rewrite in C++ will allow easy implementation of handling multiple input devices and a function button.

As crazy as it sounds, I will probably have more time to develop btnx in 1.5 months when I'll have a 9-5 job developing for Symbian .

[daou]

Daou,
I've done a little digging into BlueTooth mouses, and it seems that the problem I was previously referring to can be summed up this way:
BlueTooth mouses don't show up as USB devices. The BlueTooth dongles do, but with both dongles and on-board receivers, devices are referenced only by their MAC addresses. I'm certain the input handlers would be the same, but detection would likely be problematic.

Just thought I'd share....

In that case, it would require quite a lot of changes to get bluetooth mice working. Something to keep in mind, but a low priority for now.

[daou]

blazoner:

We are pretty much in agreement. It is one thing to install btnx on this box where it is my toy and I am the administrator, but this program will never fly in a corporate environment. I work with some pretty anal retentive IT people in the telecom industry where they do not give anyone admin access unless there are all sorts of approvals. They have good reasons most of the time for their behavior.

btnx would never be installed in these enviornments unless under a generic profile that fits only one type of mouse. The architecture and security model should be reviewed with these situations in mind before it is included in any Linux distro.

Having said that, btnx is one of the most vital applications on this machine. If someone comes in my office or the phone rings, I just hit the pause button or adjust the volume. I love it!

I agree here as well. One of the strengths of Linux (and Unix) is the way it is able to handle a multiuser environment, and one where the user and administrator are two different people. btnx does not fit into this picture and it is one thing I have been wanting to fix for some time. Fixing it is definitely a high priority task, but one which needs a lot of time and planning.

[daou]

EDIT: I think command execution should only be allowed based upon the UID of the user.

Also, please consider creating a default UID that would be permanently inactive/null. This would require the user to consciously choose a good UID, not just blindly click through letting commands be executed as root.

The default null UID is a good idea, but one which will result in a lot of support questions from users. However, I believe in security before usability so this should be implemented.

Can you elaborate on why you think cmd exec should only be allowed based on UID and not user names with su?

[daou]

The rest of the posts seem to be quite old and I don't know if anyone is waiting for answers to them anymore. If you are still experiencing a problem and posted earlier, please repost with a link to the old post. I will try to help and hopefully solve any remaining issues.

[daou]

It seems that there is an error opening the btnx logo. Note the path "/usr/share/pixmaps/btnx.png/btnx.png". This is incorrect. You have just stumbled upon an at least 6 month old bug .

Interesting that no one else has had problems with it earlier. I found the bug and fixed it. It will be released with btnx-config v.0.4.8, which I will release as soon as I get the finished Russian translations from Launchpad. Alternatively, you can get the fix from the SVN repository.

Thanks for bringing this up.

btnx-config v.0.4.8 released which includes this fix.

btnx v.0.4.9 is in the SVN repo which includes:

• fix for key combo release order (affects some systems, not all)
• configuration looping: if you have multiple configurations for different mice, btnx will try to open configuration files until it finds a configured mouse plugged in. Previously it only tried one and quit if the first mouse was not plugged in.