![](https://secure.gravatar.com/avatar/cfad84812bc41579778686925f4599d5.jpg?s=120&d=mm&r=g)
1 0 0 1 115.925 622.631 cm Q 1 0 0 1 12 0 cm Q 1 0 0 1 -115.925 -622.631 cm The switched last two lines probably cause your trouble.
Of course it does, that is exactly what I'm talking about.
I wouldn't use "page" for stuff that affects the transform matrix. See also the latest pdftex.def.
This is not the question of my preferences, I'm showing a heavy inconsistence in pdfTeX behaviour. Look: 1\pdfliteral direct{q}2% \pdfliteral{q}3% \pdfliteral{Q}4% \pdfliteral direct{Q}% 5\end This gives exactly what I expect. But now just replace '4' with a space: 1\pdfliteral direct{q}2% \pdfliteral{q}3% \pdfliteral{Q} % \pdfliteral direct{Q}% 5\end and you get something completely different and hopelessly spoiled The reason is that the transformation matrix is not serialized until pdfTeX thinks its necessary. If there is no BT..ET before the second Q, the matrix is restored _after_ that operator, which leads to disaster. But I can't see the reason for that postponed matrix restore... -- Pawe/l Jackowski P.Jackowski@gust.org.pl