Use xinput to not depend on access to /dev/console (root only).

[ypid]

Closes #34

Additional improvements and fixes:

• Why use cat, grep and wc combined when grep can do all of this
  at once? Fixed. shellcheck can make you
  aware of such things.
• Never store highly sensitive information like keystrokes (which
  potentially can contain passwords) on persistent storage.
• Terminate background processes when the main process gets killed.
• Suppress bash warning when background process dies.
• Updated and fixed README.md accordingly.
• Removed trailing whitespace.

[ypid]

Please merge #32 first. I will rebase this PR then.

[csmarosi]

AFAIK when you have 2 keyboards (==laptop w. external one), you capture at most one with this.

For security purposes, I would rather pipe it to | tr -d '[0-9]' or so instead of /dev/shm

[ypid]

For security purposes, I would rather pipe it to | tr -d '[0-9]' or so instead of /dev/shm

Prefect! Thanks very much 👍 Updated my PR. I think /dev/shm should be fine or is there something against using it?

AFAIK when you have 2 keyboards (==laptop w. external one), you capture at most one with this.

You are right. Thats when I saw your PR and intended it to be merged first so that I could rebase on it and steal that peace from you. I am to lazy to check that out right now but that is a todo.

[ardentperf]

FYI, I've done a merge of @ypid and @csmarosi xinput code. On my laptop, the keyboard does not come through as device 11 -- so I needed @csmarosi code that dynamically pulls all input devices. However I also wanted @ypid code that moved the temp file to shm and stripped out scan codes.

Also, personally I'd like to include mouse clicks as well. One problem I have is that web browser activity is sometimes underrepresented at times I'm primarily using the mouse to interact rather than the keyboard. It's pretty easy to extend @csmarosi's code to include the mouse.

[ypid]

FYI, I've done a merge of @ypid and @csmarosi xinput code.

Sounds good :)

Also, personally I'd like to include mouse clicks as well.

You might want to checkout selfspy for that as it already does this. Maybe it would be worth using the logging backend of selfspy as it also supports other platforms and just looks more mature to me.

[ardentperf]

I completely agree that using selfspy for data collection would be a good idea. Much more sophisticated data collection and handling. But ulogme has the best front end - I especially love the bar codes; I can't live without them anymore. (Got hooked through Manictime on windows.) ulogme is great code!

In the meantime though, until it uses selfspy, I would vote for merging the xinput patches. I'm running the merged version that's in my GitHub repo now - seems like a good update so far. You guys wrote good patches. :)

[dcousens]

My device ID was 10... this should be configurable

[ypid]

Makes sense, done. There might be other ways to configure this for example by including a config file. I just went with environment variables for now.

[dcousens]

why not /tmp?

[ypid]

/dev/shm is almost always a tmpfs which is important for:

Never store highly sensitive information like keystrokes (which potentially can contain passwords) on persistent storage.

Ref: https://superuser.com/questions/45342/when-should-i-use-dev-shm-and-when-should-i-use-tmp

[dcousens]

@ypid I just found that the ULOGME_KEYBOARD_ID can change, without warning.
That is, suddenly my keyboard was 9, not 10 - without a restart.

Maybe using the approach used by @csmarosi in csmarosi@738196e#diff-1c19f299da97be2f85d90e241a93212bR17 is the way to do it?

[ypid]

@dcousens Looks more appropriate indeed. I have not tested ulogme recently so I am missing some experiance running it.

[dcousens]

@ypid I went with using Python instead, as succint, no root.
No process spam.