Re: [PATCH RFC] selftests: do not use .ONESHELL

From: Jiri Benc <jbenc@redhat.com>
To: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Cc: bpf@vger.kernel.org, Jiri Olsa <jolsa@redhat.com>,
	Shuah Khan <shuah@kernel.org>
Subject: Re: [PATCH RFC] selftests: do not use .ONESHELL
Date: Fri, 15 May 2020 10:28:41 +0200	[thread overview]
Message-ID: <20200515102841.3fa15ff7@redhat.com> (raw)
In-Reply-To: <20200515030051.60148-1-yauheni.kaliuta@redhat.com>

On Fri, 15 May 2020 06:00:51 +0300, Yauheni Kaliuta wrote:
> 1) I'm wondering how commit c363eb48ada5 ("selftests: fix too long
> argument") worked without the patch.

I think it was because it reduced the list of files from three
replications to two. I did not notice the .ONESHELL; it also explains
the oddity that I saw with @ behavior.

With the .ONESHELL removed, we can further simplify INSTALL_SINGLE_RULE
by removing the @echo rsync and the at-sign before rsync.

> 2) The code does not look working as expected for me:
> 2.1) "X$(TEST_PROGS) $(TEST_PROGS_EXTENDED) $(TEST_FILES)" != "X" is
> always true sine the left part will be at least "X  " (spaces);
> 2.2) according to manual in .ONESHELL case gmake checks only first
> line for @, so @rsync is passed to the shell;
> 2.3) $(OUTPUT)/(TEST_PROGS) adds $(OUTPUT) only to the first prog;
> 
> Did I miss something?

I think you didn't miss anything and that you're right. Could you
submit a patch to remove the spaces? I can then submit a patch to
further simplify INSTALL_SINGLE_RULE if you don't want to do that, too.

Thanks!

 Jiri