2004.07.11 17:56 "Re: [Tiff] unintentional ABI change between 3.5 and 3.6?", by Jay Berkenbilt

To add to my previous statements, I wonder why TIFFRGBAImage isn't an incomplete type like TIFF. Also, are there other structures whose details are inadvertently exposed to the public interface? I don't see any. Or have I missed the boat on this analysis?

Maybe the right fix is to make TIFFRGBAImage an incomplete type and then increment the soname and call this 4.0.0 (or separate the soname from the library version). Then future changes to _TIFFRGBAImage won't create ABI changes.

--Jay

Jay Berkenbilt <ejb@ql.org>
http://www.ql.org/q/