fix(gdb): fix regression that fails to generate command names

[akinomyoga]

The current version fails to generate any command names for the first word after the command name "gdb". This is a regression introduced in #1086 and pointed out in the comment: 73c5292#r138406647.

[scop]

Thanks!

[akinomyoga]

Thank you! However, the CI doesn't seem to pass actually. In making the test, I assumed that awk would be always found in PATH, but it doesn't seem to work in the alpine container. I'll later take a look.

[akinomyoga]

but it doesn't seem to work in the alpine container. I'll later take a look.

bash-completion/completions/gdb

Lines 30 to 34 in ca8e240

	_comp_compgen_split -o plusdirs -- "$(
	find ${path_array[@]+"${path_array[@]}"} . -mindepth 1 \
	-maxdepth 1 -not -type d -executable -printf '%f\n' \
	2>/dev/null
	)"

The find command in the alpine container doesn't have the operator -printf. According to POSIX [1], -mindepth is not POSIX, -maxdepth is not POSIX, -not is not POSIX, -executable is not POSIX, and -printf is not POSIX. Only -type out of the six operators used in this implementation is POSIX-compatible.

• [1] XCU 4 - find

It didn't cause an error before because it's never been tested.

[scop]

I suppose we could use require_cmd=True and gdb as the test command.

[akinomyoga]

I suppose we could use require_cmd=True and gdb as the test command.

Ah, yeah. That is clever.