Re: [PATCH 2/5] ci/lib: allow running in GitHub Actions

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Apr 12, 2020, 21:42 UTC
Message-ID: <nycvar.QRO.7.76.6.2004122339290.46@tvgsbejvaqbjf.bet>
In-Reply-To: <xmqq7dynjgx7.fsf@gitster.c.googlers.com>

Hi,

On Fri, 10 Apr 2020, Junio C Hamano wrote:

Show 18 quoted lines

> SZEDER Gábor <szeder.dev@gmail.com> writes:
>
> >> > CI_OS_NAME=osx.  This is head-scratchingly backwards, and I think
> >> >
> >> >   test "$CI_OS_NAME" = macos && CI_OS_NAME=osx
> >> >
> >> > would read better.
> >>
> >> I can understand where you come from, but your code is not `set -e` safe,
> >
> > It works as expected in at least in dash, Bash, BusyBox sh, ksh,
> > ksh93, mksh/lksh, yash, posh, FreeBSD /bin/sh, NetBSD /bin/sh.
>
> Thanks for a clarification.  I do not use "set -e" myself (it is not
> a good idea to rely on it in general), and was wondering if what was
> said up there was true, as it did not sound like a useful behaviour
> at all.  Not complaining about a non-zero exit before && or || makes
> it usable ;-)

While it is not in general useful to rely on `set -e`, there are situations when it comes in handy for debugging. Or in copy-edited code.

As to the "It works as expected", this is in a Bash:

	$ test "$CI_OS_NAME" = macos && CI_OS_NAME=osx

	$ echo $?
	1

Yes, it works, but in the common case, it exits with an error (that we ignore, unless run with `sh -e ...`).

Just making sure that we're talking about the same expectation here.

Ciao, Dscho