17 0 obj << /Type /Annot /Q 0 /Subtype /Widget /Parent 16 0 R /DA (/rmtf 11.9552 Tf 1.1955 Ts 0 0 0 rg 0 0 0 RG) /MK << /CA () >> /F 4 /Rect [0 14.4275 59.7758 26.3827]
endobj 18 0 obj << /Type /Annot /Q 0 /Subtype /Widget /Parent 16 0 R /DA (/rmtf 11.9552 Tf 1.1955 Ts 0 0 0 rg 0 0 0 RG) /MK << /CA () >> /F 4 /Rect [0 0 59.7758 11.9552]
endobj
which look like Annot's with Rect's to me. There is a parent field I was puzzled by this too, but apparently its the parent where the error message is thrown out (rightly or falsely).
Are you saying that Evince does forms (and Javascript as it's associated) or is it checking something that is not interpreted anyway? Officially a viewer should ignore annots that it cannot handle. Evince does forms by now, at least to some extent, without JavaScript of course. I've tested this: http://help.adobe.com/en_US/Acrobat/9.0/Samples/interactiveform_enabled.pdf which gives me a "Unimplemented annotation: POPPLER_ANNOT_LINE, please post a bug report in Evince bugzilla (http://bugzilla.gnome.org) with a testcase." So there must be some checking.
Anyhow, in context we've always had such widgets implemented using a parent / child model and it's the child that matters for rendering (and it definitely has a Rect as it's an Annot) while the parent is just a dictionary and not an annot (one can argue that the Subtype key is not needed there but - at least in the past - it had to be there.) Sounds fair enough, and I'm not complaining :).
So, unless I'm sure what (and if something) is wrong I'll do nothing. (Acrobat preflights the document all right and some widgets types are too instable to mess with too much anyway.) That's fine, and as long as its working I don't really care.
Hans Cheers
Adam