harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Geir Magnusson Jr <g...@pobox.com>
Subject Re: [classlib] debug compilation as default
Date Mon, 17 Jul 2006 14:46:12 GMT


Salikh Zakirov wrote:
> Mark Hindess wrote:
>> I'll let Geir speak for himself, but I think he was alluding to the use
>> of a user properties file:
>>
>>   <property file="${user.home}/.harmony.properties" />
>>
>> that would *not* be checked in and would be read (if it existed or
>> ignored otherwise).
>>
>> Which is exactly what you can achieve in a platform-independent way with
>> a property file as described above.
> 
> This solution is okay with me.

That's not a solution.  It's what people do to personalize.  I'm
targeted at normalizing how we handle configuration properties.

> 
>>> And by the way, the solution to run 'HY_CFG=xyz ant' with 
>>> corresponding 
>>>     <property environment="env"/>
>>>     <condition property='hy.cfg' value='${env.HY_CFG}'>
>>>        <isset property="env.HY_CFG"/>
>>>     </condition>
>>> worked perfectly for the DRLVM builds on Windows. What's more,
>>> it is *compatible* with 'ant -Dhy.cfg=...' syntax. 
>>>
>>> I have not heard any specific concerns why this can't be used.
>> Read the list archive.  Environment variables are platform-specific and
>> differ - if they exist - from platform to platform.  Windows for example
>> treats environment variable names as case-insensitive but when ant reads
>> them from env.VaR you must get the case exactly right.  Ant properties
>> are platform-independent and so will work the same everywhere.
> 
> I have read the discussion, and still consider this as non-issue,
> because anyone can be careful enough to define the name in correct case.
> (So one must define the variable in proper case as HY_CFG even on Windows)

The point is that we want to reduce the number of places that things can
get broken...

> 
> In my practice, environment variables worked fine on Windows.
> 
> And I have not heard a single report on how environment variables failed in real life.
> I have only seen discussion how this *may* fail if the user is not careful enough
> to do precisely what is written in instruction.
> 
> Still unconvinced why we can't allow careful users to use environment
> variables for configuration.

because as a matter of engineering practice, you want to reduce the
number of places that errors can happen, and increase the amount of
easily reusable configuration.

Notice how you keep saying "can be careful enough" or "allow careful
users"....

We can make it easier to work with Harmony, I believe, by removing the
'extra care' requirement when possible.

geir

---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
For additional commands, e-mail: harmony-dev-help@incubator.apache.org


Mime
View raw message