ant-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Bodewig <bode...@apache.org>
Subject Re: How do I cause <javac> to fail on jikes warnings?
Date Thu, 25 Oct 2001 06:14:07 GMT
On Wed, 24 Oct 2001, Conor MacNeill <conor@cortexebusiness.com.au>
wrote:

> Stefan Bodewig wrote:
> 
>> To parallel the other jikes options, you could simply introduce a
>> new magic property like build.compiler.warningsareerrors or
>> something.
> 
> Well it would be consistent, although I am not that keen on magic
> properties.

Who said I'd like it?

>> But maybe we've simply reached the point to introduce <javac
>> additionalArgs="..." />
> 
> I think this capability should probably be there

I've put it at the end of my looong TODO list, patches will get it
further to the front 8-)

> Such a build file will fail with a different compiler since the
> compiler adapter cannot know whether the arg applies to its
> particular compiler (unless it is really smart about it).

This is a common problem with all facade tasks we have (<javac>,
<rmic> and <jspc> now, more to come later) - maybe Diane's idea of
allowing flags that will only be used by the matching implementation
is the right solution here.

> It seems a balance between a mega-interface and non-portable build
> files. Thoughts about where that balance lies?

Throwing in things that depend on the implementation explicitly would
solve the problem, this is much better than bloating a facade with
attributes that only apply to one or a few implementations.  In some
way, the fork attribute is already part of such bloat.

Stefan

Mime
View raw message