dot.mkshrc robustness

[Thorsten Glaser]

Make dot.mkshrc (and prt.mkshrc) robust:

• work with “set -e” and “set -u”, often used in Debian

• not suck with “set -x”, well not much

• work with read-only filesystem, in particular no $TMPDIR (pointed out by Natureshadow)
  ‣ especially relevant for Android /system/etc/mkshrc

It's already mostly alias-proof and function-proof¹, thankfully.

This is especially relevant for both being /bin/sh (or /system/bin/sh ofc) and root's login shell, hence importance High where it'd otherwise be Medium.

① Assuming the user is not stupid and defines shell functions named “return”, “typeset”, “command”, etc. – i.e. POSIX builtins and specials plus the ksh “typeset” one.

[Elliott Hughes (enh-9)]

On Wed, Apr 8, 2015 at 2:07 PM, Thorsten Glaser
<email address hidden> wrote:
> Public bug reported:
>
> Make dot.mkshrc (and prt.mkshrc) robust:
>
> • work with “set -e” and “set -u”, often used in Debian
>
> • not suck with “set -x”, well not much
>
> • work with read-only filesystem, in particular no $TMPDIR (pointed out by Natureshadow)
> ‣ especially relevant for Android /system/etc/mkshrc

(we set TMPDIR to /data/local/tmp nowadays.)

> It's already mostly alias-proof and function-proof¹, thankfully.
>
> This is especially relevant for both being /bin/sh (or /system/bin/sh
> ofc) and root's login shell, hence importance High where it'd otherwise
> be Medium.
>
> ① Assuming the user is not stupid and defines shell functions named
> “return”, “typeset”, “command”, etc. – i.e. POSIX builtins and specials
> plus the ksh “typeset” one.
>
> ** Affects: mksh
> Importance: High
> Status: Triaged

[Thorsten Glaser (mirabilos)]

Elliott Hughes dixit:

>> ‣ especially relevant for Android /system/etc/mkshrc
>
>(we set TMPDIR to /data/local/tmp nowadays.)

Ah ok, good. Who can write there?

Still a good idea to make the rc files so robust that
the shell can at least come up cleanly… ;)

bye,
//mirabilos
--
(gnutls can also be used, but if you are compiling lynx for your own use,
there is no reason to consider using that package)
 -- Thomas E. Dickey on the Lynx mailing list, about OpenSSL

[Elliott Hughes (enh-9)]

On Wed, Apr 8, 2015 at 4:54 PM, Thorsten Glaser
<email address hidden> wrote:
> Elliott Hughes dixit:
>
>>> ‣ especially relevant for Android /system/etc/mkshrc
>>
>>(we set TMPDIR to /data/local/tmp nowadays.)
>
> Ah ok, good. Who can write there?

the shell user.

> Still a good idea to make the rc files so robust that
> the shell can at least come up cleanly… ;)

indeed.

[Thorsten Glaser (mirabilos)]

• nothing to be done for prt.mkshrc (OBS home:mirabile/mksh)
• nothing to be done for prt.mkshrc (DEB wtf-mksh)
  ‣ DEB mksh not checked
• not much to be done for dot.mkshrc
  ‣ everything done that needed doing

That leaves us with:
• fix Android mkshrc
  – quote colon arguments
  – validate TMPDIR=/data/local/tmp DWIW
  – review
• check prt.mkshrc in Debian vanilla

Both of these are not tasks inside mksh itself, but I'll track them here too.

[Thorsten Glaser (mirabilos)]

OBS and DEB wtf-mksh done, this lets Debian-vanilla (currently frozen) and Android remain

[Thorsten Glaser (mirabilos)]

Fixed everywhere but in Android, which to contribute to I don't have the resources (i.e. a VM on which it can actually built) any more due to their… interesting… build environment.