ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Conor MacNeill" <>
Subject RE: DO NOT REPLY [Bug 26020] New: - Writing to a file.
Date Mon, 12 Jan 2004 12:41:26 GMT
> From: Jose Alberto Fernandez []
> Sent: Monday, 12 January 2004 11:10 PM
> To: Ant Developers List
> For that, you really need the to be able to stop and start
> the recorder as many times as you may want (which means appending).
> Going back to the bug text, I see that the original problem
> most probably had to do with Netbeans keeping the static environment
> of ANT around and therefore ANT does not know whether this is the first
> time <record/> is called on the build or not.

Well, I think <record> has a few problems in its design which have indeed
been shown up by netbeans keeping the static map hanging around. However,
the issue is not really restricted to those environments. The example I
posted shows <record> continuing to record after it has been stopped and
appending when it not supposed to be appending.

Stefan's backward compatabilty example is not that strong a problem for me
since it relies on <record> to append on the second invocation even when
append is not specified - i.e. it relies on incorrect behaviour. If you want
to guarantee ordering in Ant, you use "depends" to ensure the file is
deleted before starting and then append in all cases.

> This is a very specific situation, applicable to those environments that
> cache ANT classes (how does this interacts with the new loader stuff
> I do not know).
> In any case, maybe what we need is a way to tell <record/> to reset
> itself
>    <record action="reset"/>

Well that makes the buildfile specific to a particular execution environment
so I do not think it is a great solution. We need to recognize that people
do run Ant within IDEs and those IDEs may not restart the JVM used to
execute Ant.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message