Re: IDL problem

On 19/09/2013 13:43, Robin Berjon wrote:
> On 18/09/2013 18:09 , Richard Ishida wrote:
>> PS: I'm really starting to wonder whether I should reconsider using
>> respec. It's so brittle!  The last three documents I've tried to publish
>> would have been published *much* faster if I'd just used a text editor.
>>   Do we have adequate tests in place, and are they used after people
>> tweak the code?
>
> No one is making you use it Richard :)

But I *want* to use it. If it works, it's great. I guess I've just been 
particularly unlucky so far.


> In this case you tried to use the output from Anolis as input to ReSpec.
> That's a bit like printing a Word document to PDF then trying to
> manipulate it as if it were HTML: you're going to have a bad time!

That's not the problem, actually. The problem was that, although you did 
much excellent work on updating the documentation for respec, I was 
unable to find something that said "You must avoid using the class name 
'idl' unless you use this structure for your data...", and had to try 
various things to work out that in fact that class name was where the 
issue originated. Even just a list of reserved class names in the 
documentation would probably be helpful. (I a problem once before with a 
different class name that produced a particular effect I wasn't 
expecting - i think it was 'req'.)

The bigger and more time-consuming problem while publishing this time, 
however, was the crash of the references db when UTS22 was added to it, 
rather than the IDL issue.

> The test suite can of course be improved but it catches a lot of issues
> already. As far as I know we've only had two or three regressions over
> the ~40 releases that shipped in the past 12 months.

The last time i tried to publish a doc we had The Case of the 
Disappearing Status Section.  That, i'd have thought, would have been 
easily apparent if the person who made the change had tried opening any 
document.  So I was wondering whether perhaps there's a process issue(?)

Sorry for the whinge. As I said, I'd like to hope that things will work 
seamlessly next time, because this should be a useful tool. And I've 
recommended it's use as standard for the i18n WG publications. I've just 
been rather unlucky lately. I hope there are at least a couple of useful 
ideas above.

KUTGW

RI




-- 
Richard Ishida

Received on Thursday, 19 September 2013 14:01:18 UTC