[gPXE-devel] [PATCH] [script] Allow control over error handling

[J.H.]

On 03/19/2010 04:24 AM, Shao Miller wrote:
> Stefan Hajnoczi wrote:
>>
>> Will this patch work for nested script execution?
>>
> 
> gPXE's current behaviour is to exit all scripts upon error.  Scripts
> developed by users might expect this behaviour and expect the regular
> fall-back mechanism.  Suppose 'autoboot' is implemented as a script some
> day or that the user uses some other top-level script which clears
> exit-on-error...  I would not want the behaviour of the lower-level
> scripts to mysteriously change without the user explicitly clearing
> exit-on-error in each script, so this patch requires one to clear the
> exit-on-error state in each script.
> 
> It might be the case that user B uses a gPXE obtained from user A, where
> user B does not know user A's embedded script, but expects classic gPXE
> exit-on-error behaviour.  I'm thinking particularly about any services
> that offer custom gPXEs, such as Boot.Kernel.Org or NetBoot.Me. 
> Someone, somewhere, could potentially be middle-manning BKO by using
> another gPXE script as "pxelinux.0".  Since BKO would likely use this
> feature, that person's middle-man script will continue to function with
> the classic behaviour.
> 
> It's certainly open to opinions, since it's easy to implement either
> way.  What do you think? :)
> 
> - Shao Miller

So I've gone ahead and pulled it into my testing tree, and it works as
expected.  I know of the general dislike of compile time options, and
such, but maybe this is worth having as a tune-able of some sort, at
least in short term.  Leave the default as-is, but make a build option
that changes the behavior so that those of us who want the more
complicated behavior can have it?  Either way I'm tempted to move
forward with using this patch as it fixes a lot of things in BKO's gPXE,
and allows us to run a much more up to date gPXE - which is quite
beneficial to us.

Just my thoughts anyway.

- John 'Warthog9' Hawley