Re: sub-process changing value available to parent process?

[Marco Ippolito]

I knew Greg would provide a great answer. This is even better than
anticipated.
Thanks!

On Tue, 28 Apr 2020 at 13:42, Greg Wooledge <address@hidden> wrote:

> On Tue, Apr 28, 2020 at 06:41:45AM +0200, Marco Ippolito wrote:
> > > > > set -euo pipefail
> > > > As soon as I saw this, I decided to delete the entire thread.  This
> > > > is not salvageable.  This is a sign of grave incompetence and/or
> > > > ignorance.  Combined with "I know the answer is eval but I want a
> > > > different answer", it points to intransigent stubbornness.
> > >
> > > Sorry I probably miss some references here.
> >
> > A reference was indeed missing.
> >
> > Among others, many of which from Greg himself, I could find:
> > https://mywiki.wooledge.org/BashFAQ/105
> >
> > Explore from that site for more.
> >
> > Greg may want to expand with additional links.
>
> Sadly, we haven't bothered to write a whole page for "set -euo pipefail"
> yet, although it probably deserves one.
>
> To really understand the issues, first you have to look at the historical
> problems surrounding set -e and set -u, both of which are well documented.
>
> https://mywiki.wooledge.org/BashFAQ/105
> https://mywiki.wooledge.org/BashFAQ/112
>
> So you've got super bad idea #1, and pretty bad idea #2, both of which
> are well known to create more problems than they solve.
>
> Now you use BOTH of them TOGETHER, because bad x bad = super bad,
> and super bad must be good!  Or something.
>
> But wait, it turns out super bad isn't terrible enough.  So, we decide
> to EXTEND set -e to make it even worse!  Normally, set -e only applies
> to the last command in each pipeline.  But what if we could make it apply
> to EVERY command in the pipeline instead?  How awful^H^H^Hesome would
> that be?!
>
> Chet gave us set -o pipefail which lets us do exactly that!  Hooray!
> Now our shell scripts can mysteriously die in the middle of the night
> for ANY reason at all, instead of just some of the reasons!
>
> Some of the people out there decided that these three things combined
> together are so incredibly bad that they should be promoted as a
> STANDARD!  They even came up with a CATCHY NAME for it!
>
> http://redsymbol.net/articles/unofficial-bash-strict-mode/
>
> Unofficial Bash Strict Mode: When you want to be absolutely certain
> that your script's chances of dying for no discernable reason have
> been fully maximized!  Accept no substitutes!
>
> This shit spread like a virus.  Everyone lost their freaking MINDS when
> they learned about this.  They decided they just HAD to get them some of
> this!  Why write scripts that would actually run, when you could play
> Russian roulette with every script?
>
> https://gist.github.com/robin-a-meade/58d60124b88b60816e8349d1e3938615
> https://github.com/carpentries-incubator/shell-extras/issues/30
> https://disconnected.systems/blog/another-bash-strict-mode/
> https://lukescott.co/2019/06/10/bash-unofficial-strict-mode/
>
> https://www.reddit.com/r/programming/comments/25y6yt/use_the_unofficial_bash_strict_mode_unless_you/
> https://news.ycombinator.com/item?id=8054440
>
> So, every newbie who has been exposed to this memetic virus has had
> their brain permanently damaged.  You see this thing cargo-culted
> EVERYWHERE.
>
> As soon as I see it in someone's question, I know to turn away immediately
> before I get even more angry.
>
> At this point, I don't know how to solve the problem that people always
> latch on to what they perceive as a Quick Miracle Cure.  We've been
> trying for years to educate people about how bad set -e is, and then
> THIS happens, and in a month, all of our work is swept away by this
> tidal wave of incompetence.
>
>