commandlinefu.com is the place to record those command-line gems that you return to again and again.
You can sign-in using OpenID credentials, or register a traditional username and password.
Subscribe to the feed for:
How to figure out what a program is doing.
-tt detailed timestamps
-f also strace any child processes
-v be very verbose, even with common structures
-o write output to file
-s N capture up to N characters of strings, rather than abbreviating with ...
This version also attaches to new processes forked by the parent apache process. That way you can trace all current and *future* apache processes.
Can be run as a script `ftrace` if my_command is substrituted with "$@"
It is useful when running a command that fails and you have the feeling it is accessing a file you are not aware of.
Locate config files of the program. May not be used for interactive programs like vim.
Useful to recover a output(stdout and stderr) "disown"ed or "nohup"ep process of other instance of ssh.
With the others options the stdout / stderr is intercepted, but only the first n chars.
This way we can recover ALL text of stdout or stderr
similar to the previous command, but with more friendly output (tested on linux)
Sometimes a program refuses to read a file and you're not sure why. You may have display_errors turned off for PHP or something. In this example, fopen('/var/www/test/foo.txt') was called but doesn't have read access to foo.txt.
Strace can tell you what went wrong. E.g., if php doesn't have read access to the file, strace will say "EACCESS (Permission denied)". Or, if the file path you gave doesn't exist, strace will say "ENOENT (No such file or directory)", etc.
This works for any program you can run from the command-line, e.g., strace python myapp.py -e open,access...
Note: the above command uses php-cli, not mod_php, which is a different SAPI with diff configs, etc.
Depending on the TERM, the terminfo version, ncurses version, etc.. you may be using a varied assortment of terminal escape codes. With this command you can easily find out exactly what is going on.. This is terminal escape zen!
( 2>&2 strace -f -F -e write -s 1000 sh -c 'echo -e "initc\nis2\ncnorm\nrmso\nsgr0" | tput -S' 2>&1 ) | grep -o '"\\[^"]*"' --color=always
Lets say you want to find out what you need to echo in order to get the text to blink..
echo -e "`tput blink`This will blink`tput sgr0` This wont"
Now you can use this function instead of calling tput (tput is much smarter for portable code because it works differently depending on the current TERM, and tput -T anyterm works too.) to turn that echo into a much faster executing code. tput queries files, opens files, etc.. but echo is very strait and narrow.
So now you can do this:
echo -e "\33[5mThis will blink\33(B\33[m This wont"
Especially for sysadmins when they don't want to waste time to add -p flag on the N processes of a processname.
In the old school, you did ;
and typing strace -f -p 456 -p 678 -p 974...
You can add -f argument to the function. That way, the function will deal with pgrep to match the command-line.
processname -f jrockit
It sits there in a loop waiting for a proccess from that user to spawn.
When it does it will attach strace to it
Traces the system calls of a program. See http://linuxhelp.blogspot.com/2006/05/strace-very-powerful-troubleshooting.html for more information.
strace can be invaluable in trying to figure out what the heck some misbehaving program is doing. There are number of useful flags to limit and control its output, and to attach to already running programs. (See also 'ltrace'.)