Bugs item #747, was opened at 2007-03-09 03:09 You can respond by visiting: http://sarovar.org/tracker/?func=detail&atid=493&aid=747&group_id=106 Category: hz Group: v1.40.1 Status: Open Resolution: Accepted Priority: 5 Submitted By: Robert (schlcht) Assigned to: The Thanh Han (hanthethanh) Summary: \letterspacefont and \pdffontexpand, again Initial Comment: When expanding a letterspaced font, the base font will be expanded, too, resulting in the error `font has been expanded with different parameters' when the base font is expanded again: \letterspacefont\lsfont\font 100 \pdffontexpand\lsfont 20 20 5 autoexpand \pdffontexpand\font 20 20 5 autoexpand Z\bye ----------------------------------------------------------------------

Comment By: The Thanh Han (hanthethanh) Date: 2007-07-08 20:08

Message: Logged In: YES user_id=710 it seems that Martin refered to stable as the stable branch in svn, not the latest stable release 1.40.3. I checked the patch again, it's good. Thanks. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-07-08 19:45 Message: Logged In: YES user_id=710 the patch bugfix747-5.diff was generated against pdftex-1.40.3 sources as it is at sarovar. I have just verified it again by downloading the tarball from sarovar, unpack it and try the patch. No complain at all, while with bug747-6.diff I got some warnings and one rejection: patching file pdftex.web Hunk #4 succeeded at 9528 (offset -1 lines). Hunk #5 succeeded at 13180 (offset -7 lines). Hunk #6 succeeded at 16751 (offset 2 lines). Hunk #7 succeeded at 17026 (offset 2 lines). Hunk #8 succeeded at 17037 (offset 2 lines). Hunk #9 succeeded at 17171 (offset 41 lines). Hunk #10 succeeded at 17193 (offset 41 lines). Hunk #11 succeeded at 17201 (offset 41 lines). Hunk #12 succeeded at 17211 (offset 41 lines). Hunk #13 FAILED at 17241. Hunk #14 succeeded at 17379 (offset 24 lines). Hunk #15 succeeded at 17644 (offset 40 lines). Hunk #16 succeeded at 19958 (offset 31 lines). Hunk #17 succeeded at 20037 (offset 31 lines). Hunk #18 succeeded at 20189 (offset 21 lines). Hunk #19 succeeded at 30880 (offset 19 lines). Hunk #20 succeeded at 31623 (offset 19 lines). 1 out of 20 hunks FAILED -- saving rejects to file pdftex.web.rej ---------------------------------------------------------------------- Comment By: Martin Schröder (oneiros) Date: 2007-07-08 17:17 Message: Logged In: YES user_id=421 The patch had one rejected hunk against stable and current. A working patch is attached; please check. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-07-04 08:13 Message: Logged In: YES user_id=710 patch updated again. Disable \pdfcopyfont for expanded and letterspaced fonts. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-07-03 21:21 Message: Logged In: YES user_id=710 I didn't consider the situation when an expanded font is copied. It seems pretty much of work is needed to get it right, so probably it is easier to disable \pdfcopyfont for expanded (and letterspaced) fonts. ---------------------------------------------------------------------- Comment By: Robert (schlcht) Date: 2007-07-03 17:39 Message: Logged In: YES user_id=2217 Thanks. Letterspacing+expansion works fine now, and \pdfcopyfont is also useful. I gather an expanded font cannot be copied again, right? (Not that I would deem this necessary.) The messages from the following are a bit weird: \tracingonline1 \font\f=cmr10 \pdfcopyfont\fA\f \pdffontexpand\fA 10 10 5 autoexpand \fA ABC \pdfcopyfont\fB\fA \fB ABC \bye ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-30 16:25 Message: Logged In: YES user_id=710 ah, sorry, here it is. ---------------------------------------------------------------------- Comment By: Robert (schlcht) Date: 2007-06-30 16:12 Message: Logged In: YES user_id=2217 the patch is missing... ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-29 15:58 Message: Logged In: YES user_id=710 patch updated again. I have tested it with the following: \font\f=cmr10 \pdfcopyfont\fA=\f \letterspacefont\fB=\fA 100 \pdffontexpand\fB 30 30 5 autoexpand \fB \input tufte.tex and it seems to work. But the following doesn't: \letterspacefont\fC=\f 100 \pdfcopyfont\fD=\fC \pdffontexpand\fD 30 30 5 autoexpand probably it's possible with some more effort to make the latter case work too, but it doesn't seem needed. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-29 13:32 Message: Logged In: YES user_id=710 patch deleted. Needs more testing. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-29 09:56 Message: Logged In: YES user_id=710 sorry, I didn't test it carefully enough. Here is another attempt (tested with the microtype manual). ---------------------------------------------------------------------- Comment By: Martin Schröder (oneiros) Date: 2007-06-29 08:52 Message: Logged In: YES user_id=421 Taco, the new_font_type -> real_font_type is from the fix for bug #722. We need an svn... :-( ---------------------------------------------------------------------- Comment By: Taco Hoekwater (taco) Date: 2007-06-29 07:22 Message: Logged In: YES user_id=1608 On my (fairly) clean sources, one hunk is rejected because of this: else begin if pdf_font_type[f] <> new_font_type then pdf_warning("font expansion", "font should be expanded before its first use", true, true); end; My pdftex.web uses pdf_font_type[f] = real_font_type in that test. I fixed the hunk, but I also get a crash on Robert's test file. ---------------------------------------------------------------------- Comment By: Robert (schlcht) Date: 2007-06-28 23:17 Message: Logged In: YES user_id=2217 Now this yields a core dump with: \pdfadjustspacing2 \pdffontexpand\font 20 20 5 autoexpand \hsize10em Lorem ipsum dolor sit amet, \bye (unless I've messed up pdftex.web with all them patches) ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-28 20:17 Message: Logged In: YES user_id=710 here is another patch, which introduces \pdfcopyfont. Example: \letterspacefont\lsfont\font 100 \pdfcopyfont\lsxfont\lsfont \pdffontexpand\lsxfont 20 20 5 autoexpand This patch also fixed a bug in expanding local font from virtual fonts: font shrinking didn't work. Big thanks to Robert for help with testing. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-28 18:42 Message: Logged In: YES user_id=710 expanding a letterspaced fonts always implies that the base font is expanded. In other words, it's impossible *not* to expand the base font when expanding a letterspaced font. thanks for further testing. It looks like a bug in checking font expansion spec. fwiw, font stretch limit is always fixed to [0..1000] and font shrink limit to [0..500]. Thus giving a negative value has the same effect like giving 0. I will play with \copyfont to see if it helps. ---------------------------------------------------------------------- Comment By: Robert (schlcht) Date: 2007-06-28 14:01 Message: Logged In: YES user_id=2217 If we can get rid of the error message, I think it can be left at that. Mind you, the problem is not how to expand a letterspaced font but how to *not* expand the base font when expanding a letterspaced font. That expanding a letterspaced font results in expansion of the base font can be documented as a feature, IMO. Here are some more clues: The error message seems to be related to the sign of the stretch parameter. First of all, with \letterspacefont\lsfont\font 100 \pdffontexpand\lsfont 20 20 5 autoexpand the base font will be expanded, however it will never be shrunk. Then, \letterspacefont\lsfont\font 100 \pdffontexpand\lsfont 20 20 5 autoexpand \pdffontexpand\font 20 -20 5 autoexpand doesn't yield an error. Neither does: \pdffontexpand\lsfont 20 -20 5 autoexpand \pdffontexpand\font 20 -20 5 autoexpand However, here comes the error again: \pdffontexpand\lsfont 20 -20 5 autoexpand \pdffontexpand\font 20 20 5 autoexpand If you expand the base font first, it's OK in any case, also with different signs. ---------------------------------------------------------------------- Comment By: Taco Hoekwater (taco) Date: 2007-06-28 07:37 Message: Logged In: YES user_id=1608 That sounds nice, I like it. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-28 06:45 Message: Logged In: YES user_id=710 I agree this is an ugly solution, which has its origin in a design mistake. However the problem here is not only how to deal with the error message, but the point is how to use font expansion with a letterspaced font. Please give it a thought so we can come up with an acceptable workaround. If we want everything to be clean, there are more things that have to be thrown away. what about this: have a primitive \copyfont which creates a copy of a font. Example: \letterspacefont\lsfont\font 100 \copyfont\lsxfont\lsfont \pdffontexpand\lsxfont 20 20 5 autoexpand ---------------------------------------------------------------------- Comment By: Taco Hoekwater (taco) Date: 2007-06-27 11:43 Message: Logged In: YES user_id=1608 If there is no clean way of solving the problem, I'd rather just document the error message as a feature. I find \pdfnewexpandfont very ugly (sorry Thanh). ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-06-24 10:46 Message: Logged In: YES user_id=710 here is a patch that introduces a new primitive \pdfnewexpandfont. Usage: \font\f=cmr10 \pdfnewexpandfont\fx \f 20 20 5 autoexpand Note that \pdfnewexpandfont generates \fx by loading a TFM corresponding to \f first, then applies font expansion in the same way like "\pdffontexpand\fx 20 20 5". So when \f is a font that doesn't have a corresponding TFM (for example, because it was generated in pdftex memory only), the above command won't work. In particular, the following doesn't work: \font\f=cmr10 \letterspacefont\lsfont\f 100 \pdfnewexpandfont\fx \lsfont 20 20 5 autoexpand since \lsfont doesn't have a corresponding TFM but was generated by \letterspacefont. To create a letterspaced+expanded font, one can use: \font\f=cmr10 \letterspacefont\fx\f 100 \pdffontexpand\fx 20 20 5 autoexpand things are gettting messy... sorry about that. Testing/comments welcome. ---------------------------------------------------------------------- Comment By: Robert (schlcht) Date: 2007-04-01 00:44 Message: Logged In: YES user_id=2217 Fine for me and microtype. ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-03-31 09:52 Message: Logged In: YES user_id=710 either name is ok with me. We need a new pritimive to load a new font and expand it, since the current one doesn't load a new font and hence is making trouble. I would wait for a while to see how Robert sees it before making any change. ---------------------------------------------------------------------- Comment By: Nobody (None) Date: 2007-03-19 11:07 Message: Logged In: NO Hello, I do not think that \pdfexpandfont is a good name for a new primitive, as it is too easy to confuse with \pdffontexpand. What about \pdfexpandnewfont, as -- if I understand correctly -- a new font is loaded? Jonathan P.S: Thank you for creating pdfTeX! ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-03-10 20:18 Message: Logged In: YES user_id=710 Phil Taylor has a suggestion that seems great to me: introduce a new primitive \pdfexpandfont instead. How do you see it? ---------------------------------------------------------------------- Comment By: Martin Schröder (oneiros) Date: 2007-03-10 19:24 Message: Logged In: YES user_id=421 Forget that comment. Of course a change of \pdffontexpand has to be considered carefully. By you. :-) ---------------------------------------------------------------------- Comment By: Martin Schröder (oneiros) Date: 2007-03-10 19:22 Message: Logged In: YES user_id=421 No. It's an undocumented primitive. ;-) ---------------------------------------------------------------------- Comment By: The Thanh Han (hanthethanh) Date: 2007-03-09 17:28 Message: Logged In: YES user_id=710 sigh. A mistake when implementing the \pdffontexpand primitive: it should load a new font for expansion, instead of expanding a loaded font. Too late for change now? ---------------------------------------------------------------------- You can respond by visiting: http://sarovar.org/tracker/?func=detail&atid=493&aid=747&group_id=106