A universal command for quickly checking potential errors and highlighting interesting parts in any log.
This tool mimics how humans read and identify error- or bug-related keywords in statements. It highlights these keywords, enabling users to spot them at a glance.
The selected keywords, which relate to errors, bugs, and other important or interesting elements, are based on my experience reading computer logs.
adb logcat:
syslog:
dmesg:
gpu:
python:
apt:
vpn:
strace & wget:
Make 2 commands, erra (i.e. shows "a"ll lines) and err:
$ git clone https://github.com/limkokhole/Catch-Err-All
$ cd Catch-Err-All
$ chmod +x erra
$ chmod +x err
$ sudo ln -s "$PWD"/err /usr/bin/err
$ sudo ln -s "$PWD"/erra /usr/bin/erra
[1] Similar to grep, i.e. err /var/log/syslog, adb logcat | err, tail -f /var/log/syslog | err
[2] erra to show all lines, such as dmesg | erra, adb logcat | erra.
[3] Based on above setup, you can also manually type -all option when using err, without require you edit left err to erra.
[4] This command is wrapper of grep --color=auto -niP, you can add grep option as usual, such as -r.
[5] In addition erra, use -A/-B/-C grep options with err(not erra) to limit the lines.
[6] Use err -v to view non-match lines, i.e. normal log.
[7] Since default already -n, you can use -nn (i.e. no line number) option on err or erra to hide line number.
[8] Do foo_command | grep --line-buffered bar | erra OR foo_command | unbuffer -p other_command bar | erra if want to add extra grep or command.
[1] Patterns such as "err" covers "errno" and "interrupt" but I separate them to easier distinguish in a glance.
[2] To avoid internal PCRE error: 0 error, I need regex () as few as possible. You can check if your regex exceed default maximum by echo malware | err, it will produces internal PCRE error: 0 error if regex too long. I noticed adding more entries in existing (a|b|c|...) and more [-]? doesn't causes this error.
[3] The negate style is in the format:
(?<!negate_prefix_1|negate_prefix_2)wanted(?!negate_postfix_1|negate_postfix_2)
, which the postfix don't have extra < unlike prefix.
This format useful to avoid treats common tag DEBUG as important keyword bug.
[4] (?<=\b|_) (opening boundary) and (?=\b|_) (closing boundary) acts as word boundary \b except it allow underscore _
[5] This command composes of 6 main sections divided by double \\(2 continuous lines of \), i.e. (?=\b|_) on right, (?<=\b|_) on left, (?<=\b|_) and (?=\b|_) on both side, \b\w* on both side, \b on both side, and nothing on both side. Each of them may divide into subsection by \, to easy to see prefix un/in/...etc—(many words) parts. Note that the patterns shouldn't strict to correct spelling, since log is written by anyone and need to allow incorrect spelling. And of course, nothing I can do if typo in log.
[6] The last two are "?" and "!" which normally indicate louder expression in log. But ! exclude <! html tag and [ ! shell patterns to reduce noises, while still allow != because != quite often used to express something mismatch.
[7] Between '\ , next line \ , or ', no extra space. Keep in mind don't put double |, i.e. || manually in the script which causes shows all lines.
[8] You can customize matched color of grep with e.g. export GREP_COLORS='ms=01;33'.