\sys_get_shell:nnNTF with errors defined in \keys_define

It helps doing \show\myvar and \show\testgit. The former would issue

> \myvar=macro:
->.

while the latter would issue

> \testgit=macro:
->\par .

so you see that \testgit is not empty.

By the way, you also get

|gitrev-parseHEADsh: gitrev-parseHEAD: command not found

and you need

\sys_get_shell:nnNTF {git~rev-parse~HEAD}

However the problem is that when you set \myvar you're under \ExplSyntaxOn, but not when you set \testgit. In the latter case, the system returns nothing, which generates an empty line. The \input| mechanism that's exploited by \sys_get_shell:nnNTF actually generates a pseudofile and TeX always appends another EOL at the end. So you get an empty line according to TeX tokenization rules and this is tokenized as \par.

Solution:

\ProvidesClass{test}
\ExplSyntaxOn
\keys_define:nn { test }{
    git .code = {
        \sys_get_shell:nnNTF {git~rev-parse~HEAD} {\endlinechar=-1~} \testgit {} {}
    }
}
\sys_get_shell:nnNTF {git~rev-parse~HEAD} {} \myvar {} {}
\ExplSyntaxOff
\ProcessKeyOptions[test]
\LoadClass{article}

New test.tex:

\documentclass[git]{test}
\usepackage[T1]{fontenc}

\begin{document}

|\verb|\myvar = |\texttt{\meaning\myvar}|

|\verb|\testgit = |\texttt{\meaning\testgit}|

|\ExpandArgs{e}\IfBlankTF{\myvar}{}{a}|
|\ExpandArgs{e}\IfBlankTF{\testgit}{}{b}|

\end{document}