incubator-esme-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vassil Dichev <vdic...@apache.org>
Subject Re: Wee error
Date Wed, 27 Jan 2010 12:24:00 GMT
No such thing as luck ;-)

I have actually removed the parameter for the increasing the stack
size and did a quick refactoring of the tests, the same tests seem to
fail.


On Wed, Jan 27, 2010 at 1:34 PM, Sig Rinde <sig@rinde.com> wrote:
> Well, somebody has done a good job somewhere - or been very lucky!
>
> Same environment as before. OpenJDK, Ubuntu 9.04 etc
>
> Version 903607 of esme
>
> And it built without commenting out any files!
>
> Thanks :)
>
>
>
> 2010/1/27 Richard Hirsch <hirsch.dick@gmail.com>:
>> On Wed, Jan 27, 2010 at 11:57 AM, Vassil Dichev <vdichev@apache.org> wrote:
>>
>>> I thought initially that this is some sort of infinite loop, but it seems
>>> not.
>>>
>>> The specs I suspected are "AddAction" and "AddActionNeg". Each one of
>>> these contains a monstrous for comprehension. If I delete a couple of
>>> lines from these, it works even without tuning the stack size.
>>>
>>> I think this is unnecessarily pushing the limits of the compiler. In
>>> the first part of the for comprehensions every line depends on the
>>> previous, so the compiler has to check every new line in the context
>>> of new variables bound by the previous lines.
>>>
>>> I don't think every new test line depends on the others, right?
>>> Probably this is the case for login and pools, but otherwise the
>>> compile will be unnecessarily slow, even when it manages to build.
>>>
>>> Can we refactor the test based on actual dependencies in the execution
>>> of API calls?
>>
>>
>> Of course. Probably easier to have multiple small tests rather than one
>> large test. You can probably comment the two tests and I'll break it down
>> into smaller tests later.
>>
>>
>>>
>>> Vassil
>>>
>>
>

Mime
View raw message