[HACKYSTAT-DEV-L:161] Re: Preliminary result Re: Re: Hackystat goes on diet: drops 10 KLOC in 24 hours!

(Cedric) Qin ZHANG Mon, 11 Oct 2004 02:31:38 -0700

An interesting note is that I have never seen this problem on a single-CPU box.

Increasing sleep time only hides the problem.

Mike, you might want to scan your code and see whether you have explicitly flushed and closed all output streams before exiting locc task. This might solve the problem (my guess).

Cheers,

Cedric

Michael Paulding wrote:

Hongbing,
I think I may have identified the culprit.  In the build.utils.xml file
in the hackyBuild module, the following occurs in the locc task:
<locc sizetype="javaline"
      outfile="${module.name}.size.xml"
      outdir="${locc.report.dir}">
  <fileset dir="${module.src.dir}" includes="**/*.java"/>
</locc>

<sleep seconds="2"/>

<antcall target="sendLoccDataToHackystat"/> </target>
I have a very strong suspicion that the "sleep" delay is not sufficient
at certain times during the nightly build.  Basically, if the XML output
file has not been fully written, the LOCC sensor will fail when sending
metrics for whatever module it is working on.  That would explain the
erratic behavior of certain modules not having LOCC data.
I increased the delay from 2 to 30 seconds as an initial test and will
keep an eye on the Ant log for the next week or so to make sure it
stabilizes.
Best regards,
Mike
----- Original Message -----
From: Hongbing Kou <[EMAIL PROTECTED]>
Date: Friday, October 8, 2004 5:21 pm
Subject: Re: [HACKYSTAT-DEV-L:147] Preliminary result Re: Re: Hackystat
  goes on diet: drops 10 KLOC in 24 hours!
Hi, Mike,
In configuration instance-ALL there is no data on hackyTelemetry sent out to server on day Oct 7, 2004. My suspicious is on locc sensor. Can you check the ANT LOCC sensor to see what's going on? I will do the same thing to see whether there is any bug too.
Thanks,
Hongbing
At 11:38 AM 10/8/2004, Hongbing Kou wrote:
Hi,
My preliminary is that the LOCC sensor works fine and sent data
to public

server. I already identified where the problem is from. Please look at the attached
locc.xml

for details. Three modules were missing in Oct 6 and one was added in Oct 7. I think
there is
some work to
be done in DailyProjectFileMetric. I will work on it.
Thanks,
Hongbing
At 10:24 AM 10/8/2004, (Cedric) Qin ZHANG wrote:
Please go to "extra" page, choose "Telemetry Direct Streams"
analysis,>>and input the following:
streams SLOC = {FileMetric("SLOC")};
streams SLOC = {WorkspaceFileMetric("SLOC")};
There is no FileMetric data for hackyTelemetry on OCT 7. I don't
know why
this happened.
Cheers,
Cedric
Philip Johnson wrote:
I'm getting Project Daily Summary reports, and the last two
days show an
anomaly.
Here's the October 6 report:
hacky2004-all on 06-Oct-2004 Active Time: 3.83 (total), 0 (yours) Build: 4 (total), 4 (successful). Churn: 898|162 (Total Added|Deleted), 235|0 (yours) Commits: 30 (Total|Distinct), 1 (yours) Coverage: 81.4% (Overall), 2574 (covered methods), 589
(uncovered>>>>   methods)
 File Metrics: 856 (classes), 3920 (methods), 56982 (LOC)
 Performance: 26 (total tests), 4 (failures)
 Review Active Time: (No data summary method available.)
 Review Issues: (No data summary method available.)
 Unit Tests: 532|5|1 (Total p|f|e),
Note that the total size is around 56 KLOC.
Now look at last night's report:
hacky2004-all on 07-Oct-2004 Active Time: 10.08 (total), 3.67 (yours) Build: 4 (total), 4 (successful). Churn: 1484|400 (Total Added|Deleted), 0|0 (yours) Commits: 78 (Total|Distinct), 0 (yours) Coverage: 81.3% (Overall), 2582 (covered methods), 592
(uncovered>>>>methods)
 File Metrics: 754 (classes), 3409 (methods), 47934 (LOC)
 Performance: 26 (total tests), 2 (failures)
 Unit Tests: 612|10|2 (Total p|f|e), 11|0|1 (yours)
The total size has dropped by almost 10K, even though the Churn
data

indicates there should be a net increase of 1K to the code! It
almost

looks like a module has been taken out of hackystat-ALL or
something.>>>Does anyone have any insights into what's going on?
Cheers, Philip p.s. I continue to hack away on hackyReview. I've rewritten a
lot of the

package as I've discovered what I think are some interesting
design

patterns for the DailyProjectData level. I hope to commit in
the next
few days.

[HACKYSTAT-DEV-L:161] Re: Preliminary result Re: Re: Hackystat goes on diet: drops 10 KLOC in 24 hours!

Reply via email to