Re: [PATCH] test-lib: catch misspelt 'test_expect_successo'

From: Jeff King <peff@peff.net>
Date: Mar 25, 2026, 05:46 UTC
Message-ID: <20260325054601.GA3701549@coredump.intra.peff.net>
In-Reply-To: <xmqqy0jhxb3r.fsf@gitster.g>

On Tue, Mar 24, 2026 at 12:48:24PM -0700, Junio C Hamano wrote:

Show 9 quoted lines

> > Using "-e" makes me very nervous, given all of its quirks. Granted, most
> > of them are related to it _not_ kicking in when you'd want it to, but I
> > worry it will create false positive/negative headaches.
> 
> After looking at a few scripts, I am not suffering from such
> headaches yet; it does not look too bad.  I'll stop this effort for
> now, but with a handful of patches I already sent, more than 80-90%
> of the entire test scripts that I run are now "set -e" clean, I
> think.  Note that I do not run svn, cvs, or p4 tests ;-)

Clean in the sense that you don't _notice_ any problems. But there may be lurking ones. For example, given this:

  set -e
  foo() {
	false
	echo foo
  }

what would you expect the output to be for:

  echo before &&
  foo &&
  echo after

versus:

  echo before &&
  foo

Whether that "false" triggers "-e" depends on where in the &&-chain the call to the containing function is. So things that are not problems now may suddenly become ones when far-away code is changed.

Maybe it's enough that people would notice and debug them when they happen (if "set -e" is in test-lib.sh), and they wouldn't come up all that much. I dunno. I just have been bitten enough by "-e" quirks that I'm wary.

-Peff