Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [wtp-dev] Some Updated Info on our "Build Machine" (and JUnit failures)


Hi David,

I don't see the harm in testing with IBM's VM. I think it's a good idea. I think we should test with as many flavours of VM as possible. (There are differences between the various VMs and I'd really like WTP to be able to catch failures with specific VMs using our automated bucket.)

I didn't really mean to propose an alternative. I just meant to make some comments on the move to the Eclipse build machine. However, as an alternative until bug 122833 is resolved, if our old build machine is still available, can we continue to run weekly I-builds on that machine as well so the WTP test bucket can be run on linux with the Sun VM? (Feel free to shoot this down if it involves a lot of work.)


Lawrence Mandel

Software Developer
IBM Rational Software
Phone: 905 - 413 - 3814   Fax: 905 - 413 - 4920
lmandel@xxxxxxxxxx



David M Williams <david_williams@xxxxxxxxxx>
Sent by: wtp-dev-bounces@xxxxxxxxxxx

02/06/2006 12:51 PM

Please respond to
"General discussion of project-wide or architectural issues."

To
"General discussion of project-wide or architectural issues." <wtp-dev@xxxxxxxxxxx>
cc
Subject
Re: [wtp-dev] Some Updated Info on our "Build Machine"        (and        JUnit        failures)






Thanks Lawrence, you make some excellent points.


And, yes, we should run the tests are more platforms. (I don't see harm in running with IBM's VM, since .... we want to support that one too).
Perhaps someone could try IBM's VM on windows, and see if the same error's occur?  (that is, to help narrow down if a VM difference, or a true platform difference).
[I know my particular tests would have failed on windows too, purely a VM specific issue, but don't know about others.]


I'll let Naci decide our long term plan and strategy, but ... it wasn't clear, to me, exactly, what alternative you were proposing?  


Thanks much,








Lawrence Mandel <lmandel@xxxxxxxxxx>
Sent by: wtp-dev-bounces@xxxxxxxxxxx

02/06/2006 12:11 PM

Please respond to
"General discussion of project-wide or architectural issues." <wtp-dev@xxxxxxxxxxx>

To
"General discussion of project-wide or architectural issues." <wtp-dev@xxxxxxxxxxx>
cc
Subject
Re: [wtp-dev] Some Updated Info on our "Build Machine" (and        JUnit        failures)








While the move to the Eclipse build machine should speed up our builds (good) I don't think running the tests on this machine will be a plus for us. I'd assume most of our users are not running on PPC machines and that the most popular VM in use with WTP (or Eclipse) is from Sun, which means we are not really testing in an environment that simulates that of our users. I think bug
122833 is already open for this. (Jeff - any progress?)

"And, I suspect will be even harder for any of you to reproduce locally."


I also don't think this is a good position for WTP developers to be in moving forward. We have enough work to do without trying to understand the differences between our local configuration and the build machine that are causing test failures.


Lawrence Mandel

Software Developer
IBM Rational Software
Phone: 905 - 413 - 3814   Fax: 905 - 413 - 4920
lmandel@xxxxxxxxxx

David M Williams <david_williams@xxxxxxxxxx>
Sent by: wtp-dev-bounces@xxxxxxxxxxx

02/06/2006 01:50 AM

Please respond to
"General discussion of project-wide or architectural issues."


To
<wtp-dev@xxxxxxxxxxx>
cc
Subject
[wtp-dev] Some Updated Info on our "Build Machine" (and JUnit        failures)











Some of you may notice a few JUnit failures on the latest M-builds and I-builds. (currently 4 failures).

I *think* these are all probably due to us changing the VM that runs the tests. And, I suspect will be
even harder for any of you to reproduce locally.

The reason we changed VM's that run the tests, is that we've changed the machine we are running
the builds on. It is now an eclipse.org owned machine with a PPC processor. I am the last person
who knows what that really means, but according to our friendly webmaster:
"an IBM Power5 box with four PPC CPUs and 16 GB

of RAM. The /home filesystem (downloads, cvs, etc) is connected via Gigabit

to the eclipse.org backend.  

"


And, to make matters worse, these builds are not quite sending out notifications correctly ... I suspect
we'll have fixed on Monday, but if not, please check frequently "for the latest build".

(And, to make matters worse, Naci's hard drive crashed and he has limited email :(

So .... the point is, if you can fix the tests to work on any VM, that's great .. but if you agree they are
small quirks with the tests just do to the VM, then you might remove them for now (and open
a bug to fix later).

Of course, ... maybe the failures show something important, that we should fix right away, since we
want to support all platforms ... but ... thought I'd send this hint that it might be due to the VM changing.
(I had some encoding ones that were known to be VM sensitive, so I just removed them for now).

If any questions or concerns, you know where to ask :)

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[do not read further unless you enjoy nitty gritty details ... is mostly for long term reference,
for complete reproducibility, if ever needed]


So, for "full reference" information,

we are using the 32 bit VM at available at


https://www14.software.ibm.com/webapp/iwm/web/reg/download.do?source=lxdk&S_PKG=ip32142sr4&cp=UTF-8


(you might need to minimally register for that url, sorry, not sure of other sources of PPC 32/63 bit VM's).

(We tried using the 64 bit VM, but had to "back off" since, SWT only works with the 32 bit version of GTK)


the -fullversion options shows
J2RE 1.4.2 IBM build cxppc32142-20060120 (SR4)


There is a "buildmachineinfo.properties" file produced with each build and uploaded to download site, but isn't currently
"published" on build page ... but if you ssh or sftp to the download site, you'd see the lomboz server produced builds
are identified with buildComputer="buildserver" and the eclipse.org machine is identified
as buildComputer="utils" (which is why I'm documenting them here :) .

                 
_______________________________________________
wtp-dev mailing list
wtp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-dev
_______________________________________________
wtp-dev mailing list
wtp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-dev

_______________________________________________
wtp-dev mailing list
wtp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-dev


Back to the top