Re: [platform-text-dev] [platform-ui-dev] [proposal] Unite platform text

["David M Williams" <david_williams@xxxxxxxxxx>]

I just wanted to ask for a clarification ... and make
a comment. I do not intend to "argue" about any of what you have
said. But, I did not want to leave anyone with the wrong impression (or,
at least, leave them with only one impression ... there are probably multiple,
valid points of view). 

> The change was made in early September and the
problems reported in bug 478984 only appeared *one month later*, due to
a latent defect in our test infrastructure.

I do not know what you mean by "a latent defect
in our test infrastructure"? Do you mean an existing unit test should
have caught the error? If so, I'm wondering why. 
But, suspect you mean something else. If you did mean
an unit test should have caught it ... well, that is another great way
to contribute to Eclipse! 
The thing that "caught" the error was in
our build and happened not due to any bug, but because Tycho "eclipserun"
plugin creates a "minimal runtime". Apparently not all "minimal
runtimes" demonstrate the error, but some do. So, would be hard for
any pre-existing test to have caught it. 

> and we’ve driven away a contributor

I don't see that at all. If Alex decided to not contribute
any more, I don't think the reason was because he was "driven away".
I thought he got some good comments and suggestions, and up to him if he
wanted to continue. But, obviously, we did not want to leave in a change
that broke our build, and that of others (WTP later hit the same problem).
To some extent, it was "good luck" that it broke our build since
otherwise it might have broken other "minimal runtimes" and not
have been discovered for many more months. 

ContentType, which is used to decide what encode/decoding
rules to apply to a text file, is very low level. It is required by any
well behaved application or plugin that reads a file  and it was hard
to implement in the first place (especially to be quick and efficient).
Its initial implementation probably took a "man year", so I don't
think the fact that it is "hard to change", means that it is
brittle. To me, it just means it is complicated.  

Again, I don't mean to argue, or say you are "wrong"
in your perceptions. I simply wanted to say there are other valid ways
to perceive it, and did not want readers of this list to be left with only
one impression. 

But most of all, I wanted to understand better what
you meant by "due to a latent defect in our test infrastructure"?

Thank you, 







From:      
 Brian de Alwis <briandealwis@xxxxxxxxx>
To:      
 Eclipse Platform Text
component developers list <platform-text-dev@xxxxxxxxxxx>, 
Date:      
 10/13/2015 12:47 AM
Subject:    
   Re: [platform-text-dev]
[platform-ui-dev] [proposal] Unite platform        text
and UI projects forces
Sent by:    
   platform-text-dev-bounces@xxxxxxxxxxx

---

On 10-Oct-2015, at 9:14 PM, Pascal Rapicault <pascal@xxxxxxxxxxxx>
wrote:
> Seeing what happened with the recent changes in the content type bundle,
where patches have been integrated w/o much background is just worrisome
to me.


Alex’s change improved overall code quality: at its essence, it turned
a static singleton, which was inappropriately accessed from a number of
places, into an OSGi service.  The change was made in early September
and the problems reported in bug 478984 only appeared *one month later*,
due to a latent defect in our test infrastructure.  So now we’re
back to the original crufty code, the same faulty test infrastructure,
and we’ve driven away a contributor.

That a committer needs intimate knowledge of decisions made years ago means
our code is incredibly brittle. And that’s not a good situation to be
in.  The only way out of that problem is to start breaking it and
building tests to encode that knowledge.

One of the positive reasons to support merging the projects is that it
does broaden involvement in the project. Right now I don’t bother looking
a Core Resources and Platform Text bugs: it’s just not worth my time to
try to shepherd patches, but as a result I also don’t spend any time trying
to understand and improve those code bases.

Brian.

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