[PATCH] API review of KCalendarSystem* and KDate* widgets

Andreas Hartmetz ahartmetz at gmail.com
Fri Jul 20 00:30:55 BST 2007


On Thursday 19 July 2007 21:03:53 John Layt wrote:

>  * The yearString, monthString, dayString methods take a bool to indicate
>    using short or long date format, these should have been replaced with
>    enums/flags per API guidelines, but an lxr scan showed several hundred
>    usages, too many to fix now.  (Removed enum someone had declared without
>    actually using it.)
>
Well, that does not sound nice. People will need to write calls like
dayString(dateFormat == fancyDateFormat)
which I'd call unnecessary manual conversion of types, and it goes against 
the "no bools in API" policy.

You could, for example, deprecate the old methods and implement the new API 
and then start "porting", eventually removing the old API. There is a chance 
that this will pass as a necessary evil BIC change. Or you leave out 
the "removing deprecated API" part.
Another, less convenient, option would be to recruit some people on this 
mailing list to port some module each. The work should be really simple and 
need no more than an hour or so per module, I guess.
I volunteer for "something" if it's less than 100 calls to port :)
lxr apparently isn't perfect, and I prefer to check out most of svn.kde.org 
and do a grep -R to find the calls in such a situation. Grep to a file, look 
for false positives and filter them out etc.

> Testing:
>  * Ran exisitng unit tests and some new, all passed
>  * The existing unit tests do not have full coverage of the existing code
> or the new code, I'm in the process of writing a full set.
> [...]
> Planned steps from here:
>  * Unit tests for new methods (planned before full freeze applies)

Great! That's the spirit :)


-- 
- This place reeks of adventure and excitement, Sam!
- I thought it was this tuna fish sandwich I found crawling with life in my 
  coat pocket.




More information about the kde-core-devel mailing list