this could be done by git log -1 --format=%as | sed 's/-//g'.

also, the clone would have to happen before the version gets generated.

I post this issue (instead of creating a PullRequest) because GitHub did not allow me to create a branch in this project using my newly-created PersonalAccessToken:

$     git push --set-upstream origin jreiser-zerocopy
Username for 'https://github.com': [email protected]
Password for 'https://[email protected]@github.com': 
remote: Permission to alexmyczko/autoexec.bat.git denied to jreiser.
fatal: unable to access 'https://github.com/alexmyczko/autoexec.bat.git/': The requested URL returned error: 403

Anyway:

$ git diff
diff --git a/bashx b/bashx
index ff9cbdf..0f922a6 100644
--- a/bashx
+++ b/bashx
@@ -2,6 +2,6 @@
 # bash compressor
 # upx but for bash scripts only
 echo '#!/bin/sh' > ${1}xz
-echo 'sed 1,2d "$0"|unxz|bash -s -- "$@";exit' >> ${1}xz
-xz -c -9 $1 >> ${1}xz
+echo '(sed -n 2q; exec unxz)<"$0"|bash -s -- "$@";exit' >> ${1}xz
+xz -c -6 $1 >> ${1}xz
 chmod +x ${1}xz

The main tweak is for sed to quit after positioning the input file (and have unxz read the rest of the file directly) instead of sed copying the rest of the input into a pipe which then is read by unxz. Secondary tweaks: unxz is executed by /bin/sh which might not be bash, and minimal shells might not implement default tail merging (avoiding clone) of the last process in a pipeline; so use exec to force the tail merge. Also, the documentation for man unxzsays that compression effort -7 through -9 are likely a waste of time and RAM, especially for de-compression, where they require 17MB, 33MB, or 65MB instead of only the 9MB for -6.