On 7/30/12 2:36 PM, Ariel Constenla-Haile wrote: > On Mon, Jul 30, 2012 at 10:08:05AM +0200, Jürgen Schmidt wrote: >> On 7/28/12 11:57 PM, Ariel Constenla-Haile wrote: >>> Hi Regina, >>> >>> On Sat, Jul 28, 2012 at 6:41 AM, Regina Henschel >>> <rb.hensc...@t-online.de> wrote: >>>> Hi, >>>> >>>> I start to build trunk with MSVC Express on WinXP. The build breaks in >>>> qadevOOo with error message >>>> >>>> C:\AOO_2012_07_git\trunk\main\qadevOOo\runner\util\ValueChanger.java:378: >>>> cannot find symbol >>>> symbol : class TextFitToSizeType >>>> location: package com.sun.star.drawing >>>> if (oldValue instanceof com.sun.star.drawing.TextFitToSizeType){ >>>> ^ >>>> Reason seems to be, that the changes in >>>> r1366087: #116001# TextToFitSize item/property optimized to boolean >>>> are incomplete. >>> >>> IMHO besides of being incomplete, they are incompatible, and thus the >>> API changes should be reverted: this kind of incompatible API changes >>> should happen only on mayor versions, they should wait for AOO 4; in >>> the meantime, the solution seems to be: >> >> in general yes and I would agree but in this case Armin told me that the >> implementation was always different and the IDL was wrong. > > It doesn't matter what the implementation did internally, API users > don't care about implementation details. And it is a mistake to assume > that no API clients are using this API. In fact, it is more justified to > asume it is being used, as it's been documented in the SDK example: > main/odk/examples/DevelopersGuide/Drawing/TextDemo.java > > It also seems logical to assume that people read the documentation and > follow the examples ;) > >> We did such >> changes in the past as well but always very seldom and carefully. > > But this change is *not* carefully done: the change is not tracked in > the IDL documentation; as API user, I would expect a @deprecated tag in > the IDL, not simple removal without further information (besides the > Developer's Guide being updated too > http://wiki.services.openoffice.org/wiki/Documentation/DevGuide/Drawings/General_Drawing_Properties) > > >> It doesn't help to keep wrong IDL types that never have worked and were >> not really used. > > This is a wrong assumption, see the example from the SDK quoted above. > You cannot mesure how many API users are using this API, but assuming > that some people actually read the examples and follow them seems more > reasonable. >
I agree and apologize to not go deeper in this special case. Under this circumstance I would suggest that we revert the change and document it in IDL. The enum has no real effect and is obsolete. We can postpone this incompatible change for a 4.0 if we all agree that the change in general make sense. Juergen > > Regards >