license checker

[LudwigKnuepfer]

Compares all source files against regular expressions for licenses and creates a file structure with files named after licenses that contain a list of files that apply to the respective license. E.g:

$ ./check.sh
$ wc -l out/*
    2 out/1c-BSD
   11 out/1c-BSD-ccnl
   12 out/1c-BSD-embunit
    2 out/1c-BSD-isc
    1 out/2c-BSD-v1
    2 out/3c-BSD-atmel
    1 out/3c-BSD-clark
   26 out/3c-BSD-openbsd
    3 out/3c-BSD-ti
    1 out/3c-BSD-v1
    4 out/3c-BSD-v2
    1 out/chan
    0 out/gplv2-short-v1
   10 out/gplv2-v1
    0 out/gplv3-short-v1
    0 out/gplv3-v1
    0 out/gplv3-v2
    0 out/lgpl-short-broken-grammar
  192 out/lgpl-short-riot
    5 out/lgplv2.1-v1
   33 out/lgplv2-short-v1
   25 out/lgplv2-short-v2
   16 out/lgplv2-short-v3
    1 out/publicdomain-v1
    8 out/unknown
  356 total
$ cat out/publicdomain-v1 
boards/msba2-common/tools/src/pseudoterm.c

[LudwigKnuepfer]

Should we integrate this into Travis? I could just return 1 (indicate an error) if some unknown license is detected.

[OlegHahm]

As I pointed out during on of the recent developer meetings: I'm not in favor of such checks in the CI system, because that would blurs real building and testing issues.

[LudwigKnuepfer]

I guess someone should implement this against the github API then in order to tag the PR accordingly.

[LudwigKnuepfer]

Broken licenses are a real issue...

[thomaseichinger]

current output is

wc -l out/*
    2 out/1c-BSD
   11 out/1c-BSD-ccnl
   12 out/1c-BSD-embunit
    2 out/1c-BSD-isc
    1 out/2c-BSD-v1
    2 out/3c-BSD-atmel
    1 out/3c-BSD-clark
   26 out/3c-BSD-openbsd
    3 out/3c-BSD-ti
    1 out/3c-BSD-v1
    4 out/3c-BSD-v2
    1 out/chan
    0 out/gplv2-short-v1
   10 out/gplv2-v1
    0 out/gplv3-short-v1
    0 out/gplv3-v1
    0 out/gplv3-v2
    0 out/lgpl-short-broken-grammar
  208 out/lgpl-short-riot
    5 out/lgplv2.1-v1
   33 out/lgplv2-short-v1
   25 out/lgplv2-short-v2
   11 out/lgplv2-short-v3
    1 out/publicdomain-v1
    7 out/unknown
  366 total

shall we fix the unknown licenses to close #392?

cat out/unknown
boards/msba2-common/tools/src/boot_23xx.c
boards/msba2-common/tools/src/boot_2xxx.c
boards/msba2-common/tools/src/control_2xxx.c
cpu/arm_common/atomic.s
cpu/arm_common/common.s
cpu/lpc2387/asmfunc.s
cpu/lpc2387/startup.s

[Kijewski]

You can use ${VarName:-Default} to have a default string/fallback expression. If $VarName is undeclared or empty, then Default is used. Default can be an expression like ${VarName2:-${VarName3}}, too.
I am not sure if that is POSIX, but the plain Bourne shell and BASH support it.

[LudwigKnuepfer]

Do you think this will actually be used? I'm too lazy to think of new names right now...

[Kijewski]

With a probability bordering on certainty not.

[LudwigKnuepfer]

OK, I'm not doing it then.

[Kijewski]

I like 👍.

[LudwigKnuepfer]

I'd remove the regular expressions for license strings that have been replaced already.

[LudwigKnuepfer]

I think it's in a merge-able state now.

[LudwigKnuepfer]

squash?

[OlegHahm]

Gogogo.

[LudwigKnuepfer]

squashed & rebased.

[Kijewski]

Maybe use

cd "$( cd "$( dirname "${0}" )" && pwd )"

at the start of the script? Works in ([bd]?a|k)sh.

[LudwigKnuepfer]

Why would I want to do that?

[Kijewski]

Because a user won't expect that he has to run the script out of the script's directory.

[LudwigKnuepfer]

Why is there an extra cd $(pwd) in there?

[Kijewski]

That's the canonical example I found everywhere, I did not question the purpose of the subshell.

[LudwigKnuepfer]

A quick search only came up with DIR="$( cd "$( dirname "${0}" )" && pwd )" (which I think makes sense).

[LudwigKnuepfer]

@Kijewski how about that?

[Kijewski]

I like. 👍

[Kijewski]

ACK.