On 3 August 2012 03:14, Yanni Chiu <yanni at rogers.com> wrote:
> On 02/08/12 5:56 PM, Frank Shearar wrote:
>>
>> now get a more interesting failure:
>> http://squeakci.org/job/GitTest/9/console which shows a segfault
>> during a CompilerNotifyingTest:
>>
>> Smalltalk stack dump:
>> 0xbf8b9880 I Clipboard>clipboardText 2012164252: a(n) Clipboard
>> 0xbf8b98a0 I Clipboard class>clipboardText 2017668068: a(n) Clipboard
>> class
>> 0xbf8b98c0 I SmalltalkEditor(Editor)>clipboardText 2034626872: a(n)
>> SmalltalkEditor
>> 0xbf8b98e8 I SmalltalkEditor(TextEditor)>cut 2034626872: a(n)
>> SmalltalkEditor
>> 0xbf8b9910 I CompilerNotifyingTest>enumerateAllSelections 2033658260:
>> a(n) CompilerNotifyingTest
>> 0xbf8b9928 M CompilerNotifyingTest>testATempShadowingAnotherTemp
>> 2033658260: a(n) CompilerNotifyingTest
>> 0xbf8b9940 M CompilerNotifyingTest(TestCase)>performTest 2033658260:
>> a(n) CompilerNotifyingTest
>> 0xbf8b9958 M [] in CompilerNotifyingTest(TestCase)>runCase 2033658260:
>> a(n) CompilerNotifyingTest
>
>
> I would disable the running of CompilerNotifyingTest, by changing your
> "tests.st". I've had trouble with an image running remotely on a Ubuntu
> server - when I simply paste text into the image (via VNC), the VM
> immediately exits. I've never looked further into it, because I was mainly
> trying to debug some other problem over VNC (i.e. I avoided doing a paste,
> while debugging).
>
> Another sanity check you can do in general is to download the workspace, and
> run it locally, to make sure it runs as expected. In fact, that's usually
> step two of my usual build debugging process. Step one is to run the build
> again - sometimes, the build happens to run, just when someone is in the
> middle of uploading a set of changed packages. (I been down that road a few
> times, where I debugged for an hour, to find a bug in the code, then
> eventually discover the fix in a newer package that was uploaded around the
> time of build).

I was going to do Chris' extreme, and run just VersionNumber, but yes:
both you and Chris have given sound advice and will confirm that
everything ELSE is working. I'm reasonably sure I've broken the back
of setting up the _CI_ part. And then we can bisect until we find a
problem package, and figure out things further. It might be as simple
as needing to say -no-sound or whatever that particular switch is
called.

Another round in the CI Olympics!

frank