dash.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Denys Vlasenko <vda.linux@googlemail.com>
To: "Roberto A. Foglietta" <roberto.foglietta@gmail.com>
Cc: DASH shell mailing list <dash@vger.kernel.org>
Subject: Re: shell script error management in busybox ash
Date: Wed, 15 Sep 2021 21:47:17 +0200	[thread overview]
Message-ID: <CAK1hOcOVuR4i0FgEOF1AdTA-EbdJyOTLNbViGxT0zwHceZGt-g@mail.gmail.com> (raw)
In-Reply-To: <CAJGKYO4i85vGboyucxhJ948xZYxrBjoj5PtAO0eB3TEWW8QpzA@mail.gmail.com>

On Wed, Sep 15, 2021 at 7:48 PM Roberto A. Foglietta
<roberto.foglietta@gmail.com> wrote:
> Dear all,
>
>  Denys Vlasenko, the maintainer of BusyBox, asked me to organise a
> task to synchronise busybox ash and dash in such a way the future
> developments could benefit from each other's contributions. In
> particular this need raised after these patches has been applied:
...
>  Developing these patches we realised that the use of EXEXIT and EXEND
> exceptions complicates the code and do not add any value.

Actually there was no discussion with me to agree on such an opinion.
There may be value in it.

I didn't look too closely, but backpropagating of fatal exceptions
might be necessary to restore foreground process group on exit?
(grep for tcsetpgrp). Maybe something else too. Need to take
a very good look before deciding on "it's useless".


> Moreover,
> the code that these exceptions trigger (everywhere the setjmp()
> returns true) restores some values and frees some memory but the traps
> need the information before the restore, not the restored values.

Yes, _this_ is the problem we are having.

E.g. $FUNCNAME in EXIT trap needs to be current function's name.
With how EXEXIT exception is propagating back in current dash,
it's difficult to arrange for that: the trap is run well after
we leave the "execute this function" dash code.

The solution may be to change exit builtin handling from
"throw exraise(EXEXIT) exception, it will run EXIT trap (among other things)"
to
"run EXIT trap, then throw exraise(EXEXIT) exception
(which _will not_ run EXIT trap)"

> For
> these reason I developed three patches here in attachment:
>
>  1. get rid of EXEXIT in exitcmd() and fix a bug about return value
>  2. get rid of EXEND at the end of evaltree()
>  3. remove the EXEXIT and EXEND altogether

I would imagine if you propose changes to dash,
the patches should be against dash?


>  The description of these patches are simply arbitrary and temporary.

In other words, "my patches are not ready"... why posting them then?


> I hope we will agree on these details once we are going to port them
> to dash.
>
>  Applying these patches the output printed by trap ERR and trap EXIT
> is what was expected to be demonstrating that to avoid these
> exceptions the code became straightforward. We are here to check your
> interest in porting these patches to dash and your will to apply.
>
>  Our focus is related to the patches listed above but in the long run
> it is not limited to them but also includes the following patches.
> These patches have been applied in the meantime that others made
> changes to ash.c as well. The full list of patches and their order is
> listed here:
>
> https://git.busybox.net/busybox/log/
>
>  Other patches that have been applied that might be a convenience to
> port in dash are the following:
>
> ash: eval: Check nflag in evaltree instead of cmdloop
> https://git.busybox.net/busybox/commit/?id=41beb53787ec798a27f336c4758cb5ebd8f0c75a
>
> ash: eval: Do not cache value of eflag in evaltree
> https://git.busybox.net/busybox/commit/?id=f415e21a7dce1d4f4b760fddfaba85c551681e11
>
> ash: parser: Fix handling of empty aliases
> https://git.busybox.net/busybox/commit/?id=30af5938afad076e12b8ece123cab0b8bc92a596
>
> ash: parser: Save and restore heredoclist in expandstr
> https://git.busybox.net/busybox/commit/?id=1c06ddd8bbbd6906e5bf00ec93e04d5090718be9
>
> ash: use pgetc_eatbnl() in more places, take 3
> https://git.busybox.net/busybox/commit/?id=c54025612711a6b1997efd99206b9fbcaa5a29cf
>
> ash: parser: Fix alias expansion after heredoc or newlines
> https://git.busybox.net/busybox/commit/?id=8c68ae8416c8b54222eb3cd1d4908a570147e134
>
> ash: parser: Get rid of PEOA
> https://git.busybox.net/busybox/commit/?id=48cb983b136fb74c61db594a30e18bdc42b7264c
>
> ash: eval: Prevent recursive PS4 expansion
> https://git.busybox.net/busybox/commit/?id=eb607777697f4c5eb2dfd86e5837a8c379f65979

None of the above needs porting to dash as it already came from dash.


> ash: fix ignoreeof option
> https://git.busybox.net/busybox/commit/?id=0beee209778870888c3a80a9ae57e74888bc8e7b
>
> ash: stopped jobs should only prevent exit from interactive shell
> https://git.busybox.net/busybox/commit/?id=50239a665c88f5a95ce41146804500f5da90b19e
>
> ash: let ignoreeof only affect interactive shells
> https://git.busybox.net/busybox/commit/?id=5726df5f94f973eaa097d9853ceff2bd6b748d97

  reply	other threads:[~2021-09-15 19:47 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-15 17:47 shell script error management in busybox ash Roberto A. Foglietta
2021-09-15 19:47 ` Denys Vlasenko [this message]
2021-09-15 19:59   ` Harald van Dijk
2021-09-15 20:07     ` Denys Vlasenko
2021-09-15 20:16       ` Harald van Dijk
2021-09-15 20:22         ` Denys Vlasenko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAK1hOcOVuR4i0FgEOF1AdTA-EbdJyOTLNbViGxT0zwHceZGt-g@mail.gmail.com \
    --to=vda.linux@googlemail.com \
    --cc=dash@vger.kernel.org \
    --cc=roberto.foglietta@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).