maven-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Eggers <>
Subject Re: Maven bootstrap/surefire:test performance
Date Fri, 13 Sep 2013 19:47:50 GMT
On 9/13/2013 12:38 PM, Milos Kleint wrote:
> with Compile on Save enabled, the test-compile phase should be skipped and
> only surefire:test should be executed.
> However even with that, there's overhead of jvm startup + maven startup
> before the mojo gets executed. Obviously the overhead is biggest when you
> run just a single test. Unfortunately not much that can be done here. We've
> used to execute "Compile on Save" stuff with internal nb execution (via ant
> in IDE-jvm) but the devil is in the detail there, it's not 100% exactly the
> same execution as surefire. The current design decision is to be 100% equal
> to what cmd line executes. The downside is speed.
> Milos
> On Fri, Sep 13, 2013 at 9:25 PM, Mirko Friedenhagen <
>> wrote:
>> Hello everybody,
>> I use Netbeans 7.4 as IDE and mostly like how it uses Maven to get stuff
>> done. While it is nice that I do not encounter problems as I did with
>> Eclipse when dealing with dependency scoping I am slowed down when running
>> single test files or methods. Netbeans invokes "test-compile surefire:test
>> -Dtest=...".
>> Now even from the CLI it takes Maven some time to reach surefire:test. Even
>> when I do not invoke test-compile 3 seconds are spent before surefire:test
>> starts it's work and another 3 seconds afterwards.
>> The tests of a single testcase themselves only take 0.3 seconds. With
>> Eclipse or Intellij the execution happens in less than 1 second while with
>> Maven or Netbeans this will take up to 10 seconds even in offline mode.
>> Any hints for speeding up things would be appreciated (except of changing
>> the IDE ;-)).
>> Regards Mirko
>> --
>> Sent from my mobile

Yep, with Compile on Save and an external Maven (using 3.1) and a 
reasonably recent NetBeans 7.4 build I get about 3 seconds for executing 
a focused test method.

In short, it's not the 10 seconds you're seeing, but it's not the 0.3 
seconds you're expecting.

I can certainly live with (and in fact applaud) the design decision to 
be 100% equal to what the command line executes.

. . . . just my two cents

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

View raw message