Path: csiph.com!xmission!news.snarked.org!news.linkpendium.com!news.linkpendium.com!panix!usenet.stanford.edu!not-for-mail
From: Eli Schwartz <eschwartz@archlinux.org>
Newsgroups: gnu.bash.bug
Subject: Re: Feature Request: Custom delimeter for single quotes
Date: Sat, 2 Nov 2019 20:23:29 -0400
Lines: 113
Approved: bug-bash@gnu.org
Message-ID: <mailman.474.1572740627.13325.bug-bash@gnu.org>
References: <CAOnw=2J6fse6U=6zofMW7pORp0yTere_DYYSKGS6hf5xK2PuQA@mail.gmail.com> <13ecc4db-2b5e-95dd-2445-78191b9c01dd@iki.fi> <CAOnw=2KqGqE3zciZBqyFOBG8DxUDeCaBJUs7g2keUoKhQB0RLw@mail.gmail.com> <fe3ea34e-5af1-4e33-4107-41975dfb852b@archlinux.org> <CAOnw=2JXTMic2sW5QWRrBiBDh83kEAMVBjX2X-Qwo5jLAWh_ww@mail.gmail.com> <1b1da22e-10fd-1dd1-ce28-33e17fed0ffd@archlinux.org>
NNTP-Posting-Host: lists.gnu.org
Mime-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512; protocol="application/pgp-signature"; boundary="Ioma5PL1PITbxblKpv5tigwnLtniGH3zM"
Cc: bug-bash@gnu.org
To: Patrick Blesi <patrick.blesi@gmail.com>
Envelope-to: bug-bash@gnu.org
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.2.0
In-Reply-To: <CAOnw=2JXTMic2sW5QWRrBiBDh83kEAMVBjX2X-Qwo5jLAWh_ww@mail.gmail.com>
Precedence: list
Xref: csiph.com gnu.bash.bug:15558

This is an OpenPGP/MIME signed message (RFC 4880 and 3156)
--Ioma5PL1PITbxblKpv5tigwnLtniGH3zM
Content-Type: multipart/mixed; boundary="97dQwJTUb90G7SpVbvgF78UXe5nRi2Lbz"

--97dQwJTUb90G7SpVbvgF78UXe5nRi2Lbz
Content-Type: text/plain; charset=utf-8
Content-Language: en-US-large
Content-Transfer-Encoding: quoted-printable

On 11/2/19 12:27 PM, Patrick Blesi wrote:
> Upon further inspection, what Andreas pointed out is actually what I ne=
ed.
> Just to close the loop on everything...
>=20
> It looks like Ruby does support execution with and without a shell:
> https://apidock.com/ruby/Kernel/system.
>=20
> The reasoning for using two programming languages is that sometimes it =
is
> easier to accomplish things in Ruby and sometimes it is easier to
> accomplish things in a shell. Providing the user the option to implemen=
t
> something via shell or via Ruby allows for maximum flexibility and util=
ity.
What features of a shell are you using here? The only shell code which
you are running is:

tmux send-keys -t %1 q C-u "$command" C-m

This uses no shell syntax -- no if/while/for loops, word splitting, no
pipes, no shell builtin utilities, no process substitution, no
arithmetic expansion, and no tilde expansion. The only thing it uses is
variable expansion, but that is because you're using lots of additional
shell goop to insert a complex command into a shell variable in order to
do quotation magic for $command, instead of passing it via argv.

I am certainly not going to claim that it's wrong for ruby to include a
system() analogue. :/ I'm not even going to claim that it's wrong to
seek ways to handle single quotes in ruby's system() analogue, even
though I think it probably makes a lot more sense to use a shellescape
library for that. (General rule of thumb for shellescaping strings:
replace every instance of a single quote with the four-character
sequence: '\'' and then single-quote the whole thing. Single quotes
suppress interpretation of everything other than single quotes, and for
the single quotes themselves, you temporarily leave the quoting context
and use a backslash-escaped single quote.)

What I am going to claim is that you're incorrect for thinking you must
use a shell in the first place. Save your escaping tricks for cases
where you actually require interpreting your command in a shell, and for
this case here, use an exec. It's not just about security, which you've
already stated is not about being secured. It's not just about avoiding
erroneous handling, since your tremendous hack "probably" works, and
shellescape routines work somewhat more reliably, so avoiding erroneous
handling can be done.

It's also about efficiency: you're randomly introducing a shell in order
to run your subprocess, even though you don't need a shell at all, and
as a result your program runs slower, because it proxies subprocesses
through additional helper processes (in this case the shell).

Why are you going through convoluted, non-intuitive steps in order to
take a ruby variable and pass it as one of a series of arguments to the
"tmux" subprocess by first converting it to a shell variable using dark
sorcery and doing the passing of arguments in shell code?

Moreover since not only do you not want to interpret the user-provided
input as shell metacharacters, but you don't even want to use shell
metacharacters in your hardcoded component either.

--=20
Eli Schwartz
Arch Linux Bug Wrangler and Trusted User


--97dQwJTUb90G7SpVbvgF78UXe5nRi2Lbz--

--Ioma5PL1PITbxblKpv5tigwnLtniGH3zM
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: OpenPGP digital signature
Content-Disposition: attachment; filename="signature.asc"

-----BEGIN PGP SIGNATURE-----

iQIzBAEBCgAdFiEEvSewel70XCra9w4EhIGKaBmvSpsFAl2+HgIACgkQhIGKaBmv
SpvRrBAAmR1ANsKDDwTqvhlPZ+csvwln1QCr+T4D8Wwp7zM/Xq+qkcBgzg3pHhf7
rsd+095yw6T+huq4xwADGSsvhqM1uKcq2bIHCRwkwYob1XH51QxThKhdbNp8uHFr
A54LPlCF55l5zDX58H5KTnoQPWAmIN6gArt+rRiSJKNmv4SbWTFu2Aa0ndhJZC5s
PotKZxvDXB/gWnTaKz1qUSnRd0+bT3CxOz1ovaCZU1bBV+3npVh8yByojRD99ynX
ekaQB4gSqfJ4UjKAslN+I62VNp61JqHdvqwe+vMXgEeTaDhq8A1veLo6A+88F8SP
B69AB28vrwYcYvEVB274S4tyDyJJS1JxLmeFFuOIKgWVLBCj5oRj3dzIvYpWws62
eWaXlhINmpLNlnQvt1AO9VBYvOQ/t6C5iuFmkSZ+RgZk9ITtzKUijDaJ+Of9cUjN
cdCd2NKc6n0n+AnU0n8Rc6u0eyQgVNdsjp6luq82PuRwFoigdhqVgl2Vrxmw+aQM
KoC3fqSCinw5EpcUX26JUEdbHnaalrXLOfoAC76OI/V8LPKywbCmF/Rx6myNyG6c
nJhB8wt8luGo3wpzB6BsPy9/Mq4YflUX5YBdqkxnf5u9kuvdxkX29/eUTDWDVgFR
8GViUMosAjFvzpLliCV7nF9j+Ua0SCQdMvu2c8PkJX2irCvGh86JAjMEAQEKAB0W
IQRgQRMEwJ02YoNA7v/OsWfvtXIr1gUCXb4eAgAKCRDOsWfvtXIr1lXbEAC4iOEJ
AEX21lVxGyV9Qra3wXYQ0LxJPXZxQb50Bc5KJHigz8RLW5WL0oFYUMapXyQ4wRGq
Dic7JzNhIdp3jHAf41GQlRhmQIOKbAK5+NRsVkHsM26s/cJ1EJx59NfU43v+D25v
g9c9ABPduKvJJnKyg8DjVYQ5oSXsSRFPYpEhPmLgNLeSZnH0E40zq1tHaoU7+Ns0
aM68l6y6XdkIsdqTbxmTI0uXpgTLNS7tMtRrIjaFJOdenn+6YSF8r/Ve7QAoSWC7
lN+y85ThEGzWTGFkEXJzf3by1edAb0crv0OlG8rLzZ0cL1gULjSdz6G6Q37sVptZ
1/KoyYMLbSdBjrhUkOXyV3XpYPanWJpOeXlvyyrhAYzuGF3P1OdLXLKsM6ojYube
6Yr7ZYXOUm414HxSsD9YS+ZiXvZFc1jfzjrJ3SGjK5aaQOi/dP9EF2XNnsMcNbNX
Tn7FjWh9O0K1O9JW8MJMCnGg0pxm9f0RoqpomnIy9WcE2dO+SYgTfRDtNz/2kHo0
B43yKIaQN1/fTfwByymGzNYmUuqd/PpX9zyk+hAoxzKR6f7YUD6Zyy/TJnxkBYzP
5W4CdWxHvfTqtBlcqJT5GfYAHY4LTTx2u37pCo0O/cUGCrnhKaDe31MAX2ySV80J
mdJ/VIrawL4L0CIqB8otW0FCnNoDdbveSQ5nVg==
=Q9dF
-----END PGP SIGNATURE-----

--Ioma5PL1PITbxblKpv5tigwnLtniGH3zM--