ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Conor MacNeill" <co...@cortexebusiness.com.au>
Subject RE: cvs commit: jakarta-ant WHATSNEW
Date Tue, 30 Apr 2002 00:11:31 GMT
> From: holtdl@apache.org [mailto:holtdl@apache.org]
>
> holtdl      02/04/29 15:50:16
>
>   Modified:    docs/manual coretasklist.html tasksoverview.html
>                .        WHATSNEW
>   Added:       docs/manual/CoreTasks do.html
>                src/main/org/apache/tools/ant/taskdefs Do.java
>   Log:
>   Even though the enthusiasm (even to just vote) was a bit, ahem,
>   underwhelming, at least there weren't any overwhelming objections,
>   so...
>

-1.

Sorry for missing the discussion but I have been on leave (Anzac day holiday
in Oz).

I am quite concerned about this task. It introduces another mechanism for
running targets, in addition to <antcall>, which I'm not sure we really
need. It just seems to be such a hack to bypass the dependency processing.
The potential interactions with other "calling" mechanisms are problematic.

The implementation based on a static hashtable of target names has many
problems.

I have attached a zip with two cases illustrating how the <do> task fails
when interacting with both <antcall> and <ant>. In both cases I can execute
a target with a property not being set where you would expect, based on
dependencies, that property to be set.

Improvements could probably be made to pass these two cases but, overall, I
believe it is wrong to bypass the dependency mechanism.

Conor



Mime
View raw message