Re: [Toybox] [PATCH] Remove mount.test awk dependency.

2017-04-07 Thread Rob Landley 
On 04/07/2017 11:50 AM, enh wrote:
> Yeah, but that's a big job for the future. I am root and can spin up a
> new emulator instance and throw it away afterwards at very little cost
> (for x86/x86-64). So it's still useful in the usual "half an eye is
> better than no eye" sense. 

Oh sure, I'm just saying I want to automate it a bit more. "Warning,
dangling plumbing in this area..." :)

Rob
___
Toybox mailing list
Toybox@lists.landley.net
http://lists.landley.net/listinfo.cgi/toybox-landley.net

Re: [Toybox] [PATCH] Remove mount.test awk dependency.

2017-04-07 Thread enh 
Yeah, but that's a big job for the future. I am root and can spin up a new
emulator instance and throw it away afterwards at very little cost (for
x86/x86-64). So it's still useful in the usual "half an eye is better than
no eye" sense.

On Apr 7, 2017 09:47, "Rob Landley"  wrote:

On 04/03/2017 10:35 AM, enh wrote:
>
> Parsing file(1) output isn't a good way to determine file system type
> anyway.

The mount tests (which there need to be s many more of, there's
comments in mount.c about that) are one of the category of "tests really
needing mkroot" because:

A) Requires root to run.
B) Requires a host with known stuff on it to test
C) Hard to clean up after if it fails.
D) Can potentially hose the system badly enough to require a reboot if
it goes wrong.

The other cannonical example of this is ifconfig. :)

(Possibly it could be done with containers instead of qemu, but
insmod.test can't. And although I finally figured out how to test ps, it
requires a --bind mount over /proc after the first few simple ones.)

Rob
___
Toybox mailing list
Toybox@lists.landley.net
http://lists.landley.net/listinfo.cgi/toybox-landley.net

Re: [Toybox] [PATCH] Remove mount.test awk dependency.

2017-04-07 Thread Rob Landley 
On 04/03/2017 10:35 AM, enh wrote:
> 
> Parsing file(1) output isn't a good way to determine file system type
> anyway.

The mount tests (which there need to be s many more of, there's
comments in mount.c about that) are one of the category of "tests really
needing mkroot" because:

A) Requires root to run.
B) Requires a host with known stuff on it to test
C) Hard to clean up after if it fails.
D) Can potentially hose the system badly enough to require a reboot if
it goes wrong.

The other cannonical example of this is ifconfig. :)

(Possibly it could be done with containers instead of qemu, but
insmod.test can't. And although I finally figured out how to test ps, it
requires a --bind mount over /proc after the first few simple ones.)

Rob
___
Toybox mailing list
Toybox@lists.landley.net
http://lists.landley.net/listinfo.cgi/toybox-landley.net