On Thu, Apr 20, 2006 at 10:28:51PM -0500, Garst R. Reese wrote: > Martin Vermeer wrote:
... > >>Top Middle and Middle Middle. For this file, it works to change > >>inner_pos = ["c","t","b","s"] > >>to inner_pos = ["t","c","b","s"]. This changes the title settins to > >>Middle Top, but still gives a correct ps output. So I reverted this > >>change and used the dialog boxes to set Top Top and Middle Top and again > >>got correct output. So what dictates the ordering in inner_pos? > >> > >> > > > >Yes it happens to work, but the logic is wrong. The original order of > >inner_pos is correct and shouldn't be changed: it is the order of the > >older numerical argument inner_position. > > > > > I think the old order was a bug and this is a good time to squelch it. > in development/FORMAT lines 260-284 you have an Ovalbox and on line 269 > you use tcbs for it's inner_pos. But on line 289 you use c/t/b/s for the > Frameless box. This is just documentation... look at frontends/insets/QBox.C lines 114, 179 and you see that in the code, and the user interface, the assumed order (which in reality is irrelevant for defining the default) is "tcbs". And if you look at insets/insetbox.C line 487, you see that the default at construction is 't'. > The only place in the 1.4.1 code that I can find actually using > inner_pos is in lyx_1_2.py (line172) and lyx_1_4.py. I think my example > doc just exposed this inconsistency as a true bug. Were it not for the > strange c/t/b/s mapping all would have been well. I don't agree. You're right in that the only place where a _correspondence_ is defined between the old numerical parameters and the new character ones is, precisely, lyx2lyx. And this correspondence should be _semantically correct_, i.e. 0123 -> ctbs. (Right? You have 1.3.7 installed, so you can verify that truly 0 = center, 1 = top.) The only "bug" I see is the inconsistent default conventions: the old one height = 0, inner_position = 0, and the new one height = "1pt" (meaning, 1\totalheight) and inner_pos = 't'. The error (the ONLY error IMHO) is translating the old default to "1pt", 'c', rather than "1pt", 't'. Which my patch corrects in a minimal way. Think what your proposed remedy would do to a minipage with a manually set height, if the inner_pos semantics is wrong. - Martin
pgpEVjeRSd3v6.pgp
Description: PGP signature
