Re: [jgit-dev] Buck build for JGit

[Dave Borowitz <dborowitz@xxxxxxxxxx>]

On Thu, Dec 31, 2015 at 7:28 PM, Shawn Pearce <spearce@xxxxxxxxxxx> wrote:

On Thu, Dec 31, 2015 at 7:06 PM, Shawn Pearce <spearce@xxxxxxxxxxx> wrote:
> On Thu, Dec 31, 2015 at 6:56 PM, Shawn Pearce <spearce@xxxxxxxxxxx> wrote:
>> Also, GcCommitSelectionTest takes over 60s on this
>> machine. Buck would go faster if its long tail wasn't waiting for this
>> test.
>
> Interestingly, GcCommitSelectionTest takes 68.0s under Buck but only
> 39.269s under Maven. Something's fishy there.

I think part of it may be explained by the JVM settings Buck is using.
Its running test temporary files in the local filesystem (in a
per-test directory managed by Buck), not /tmp which is a ramdisk on
this machine (and its got tons of RAM). So we're partially waiting on
the filesystem.

If I force Buck to use /tmp for the unit tests, I can pull the from
scratch time down from 2m9.535s to 1m51.994s. But that almost just
doesn't seem worth the trouble, even though its only a 1 line change.

You can probably also improve start-to-finish runtime of tests by including more than one test class per test target, to amortize the JVM startup time.

 

_______________________________________________
jgit-dev mailing list
jgit-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/jgit-dev