Massive libxo-zation that breaks everything

Alfred Perlstein bright at mu.org
Mon Mar 2 20:01:04 UTC 2015



On 3/2/15 2:53 PM, Julian Elischer wrote:
> On 3/2/15 5:25 AM, Alfred Perlstein wrote:
>>
>>
>> On 3/2/15 4:25 AM, David Chisnall wrote:
>>> On 2 Mar 2015, at 09:16, Julian Elischer <julian at freebsd.org> wrote:
>>>> if we develop a suitable post processor with pluggable grammars, we 
>>>> save a lot of work.
>>>> given enough examples you could almost have automatically generated 
>>>> grammars.
>>> This decoupled approach is problematic.  A large part of the point 
>>> of libxo is to allow changing the human-readable output without 
>>> breaking tools that consume the output.  Now I need to keep the tool 
>>> that consumes it and the tool that produces it in sync, so that's an 
>>> extra set of moving parts.  When you throw jails with multiple 
>>> versions of world into the mix, it becomes a recipe for disaster.
> why? the jail has it own /usr/share?
>
>>>
>> +1
>
>  I think the risk is exactly opposite.  That the human readable output 
> will change subtly with bugs in the xo implementation.
> and people will not update the two output paths in exactly the same 
> way, leading bugs. I'm not going to fight on it, but I am 
> uncomfortable with it. 
So you mean that we're going to have to act like mature software devs 
and have regression tests (atf) and such?  I welcome such a change.

> You are increasing the complexity of every program you touch.
And its utility as well.  Worth it.

-Alfred




More information about the freebsd-current mailing list