Hello, ** luigi scarso [2012-02-01 02:37:20 +0100]:

On Wed, Feb 1, 2012 at 2:09 AM, Vladimir Lomov wrote:

...

Hello, please consider the following example: <example file="ex8.tex"> \starttext

...

\font\test=txr \test abs

...

\stoptext </example>

...

I process this document with context (context --version -> 2012.01.25) nothing suspecious in terminal output but in log file I see "checksum mismatch", <terminal> $ grep 'checksum' ex8.log checksum mismatch in font txr.vf ignored </terminal>

...

Simple plain tex file <example file="ex9.tex"> \font\test=txr \test abs \bye </example>

...

processed by luatex and pdftex engines (both from TeX Live 2011, I don't know how to generate formats for them in context suite) doesn't give same message.

...

Is it a "bug" in font system of context?

...

My standalone also says LuaTeX warning (file rtxptmr): Font rtxptmr at 600 not found i.e I have not the font. Maybe the tl2011 has this font ?

I specially chose that font (txr, txr.tfm and txr.vf files), it is distributed by txfonts "package". May be you installation doesn't have them because I installed context suite with './first-setup.sh --modules=all'. Nevertheless, that 'Font ... at 600 ...' reminds me how pdftex (pdflatex actually) deals with tfm/mf/pk fonts (generate pk and insert them into pdf file). --- WBR, Vladimir Lomov -- Never eat anything bigger than your head.

Hello, ** luigi scarso [2012-02-01 08:45:18 +0100]:

On Wed, Feb 1, 2012 at 7:00 AM, Vladimir Lomov wrote:

...

I specially chose that font (txr, txr.tfm and txr.vf files), it is distributed by txfonts "package". May be you installation doesn't have them because I installed context suite with './first-setup.sh --modules=all'.

...

Nevertheless, that 'Font ... at 600 ...' reminds me how pdftex (pdflatex actually) deals with tfm/mf/pk fonts (generate pk and insert them into pdf file). Even with --modules=all' LuaTeX warning (file rtxptmr): Font rtxptmr at 600 not found and $> pdffonts test.pdf

name type emb sub uni object ID ------------------------------------ ----------------- --- --- --- --------- Error: font resource is not a dictionary JFRMQG+LMRoman10-Regular CID Type 0C yes yes yes 18 0

It seems that this font is not a font of the suite.

Let me stess it in other words: this ("Font ... not found") is not the topic of thread. It is completely inrelated (IMHO) with vf font problem (vf font leads to that "checksum" message). I took that font because it has corresponding vf file. The message "Font ... not found" is on terminal output while "checksum mismatch" only in log file. Ok, IIRC, if I want to use font with plain tex (pdftex engine) I have to have TFM, and may be VF files, then MF source _or_ type1 font (pfb file), if I have only type1 font, i.e. pfb+tfm[+vf] files, I must instruct pdftex to use pfb file for corresponding tfm, this is what MAP file do, e.g. file 'original-ams-cmr.map': cmr10 CMR10

Hello, ** luigi scarso [2012-02-01 10:54:04 +0100]:

On Wed, Feb 1, 2012 at 10:21 AM, Vladimir Lomov wrote:

...

Hello, ** luigi scarso [2012-02-01 08:45:18 +0100]:

...

...

On Wed, Feb 1, 2012 at 7:00 AM, Vladimir Lomov wrote:

...

...

...

I specially chose that font (txr, txr.tfm and txr.vf files), it is distributed by txfonts "package". May be you installation doesn't have them because I installed context suite with './first-setup.sh --modules=all'.

...

...

...

Nevertheless, that 'Font ... at 600 ...' reminds me how pdftex (pdflatex actually) deals with tfm/mf/pk fonts (generate pk and insert them into pdf file). Even with --modules=all' LuaTeX warning (file rtxptmr): Font rtxptmr at 600 not found and $> pdffonts test.pdf

...

...

name                                 type              emb sub uni object ID ------------------------------------ ----------------- --- --- --- --------- Error: font resource is not a dictionary JFRMQG+LMRoman10-Regular             CID Type 0C       yes yes yes     18  0

...

...

It seems that this font is not a font of the suite.

...

Let me stess it in other words: this ("Font ... not found") is not the topic of thread. It is completely inrelated (IMHO) with vf font problem (vf font leads to that "checksum" message). I took that font because it has corresponding vf file. ok it was just to have as much data as possible.

We have

$>vftovp txr.vf (VTITLE ) (FAMILY TXR) (FACE F MRR) (CODINGSCHEME TEX TEXT) (DESIGNSIZE R 10.0) (COMMENT DESIGNSIZE IS IN POINTS) (COMMENT OTHER SIZES ARE MULTIPLES OF DESIGNSIZE) (CHECKSUM O 32212676346)

while from luatex source

source/texk/web2c/luatexdir/font/vfovf.w: @ process a local font in \.{VF} file @c static internal_font_number vf_def_font(internal_font_number f, unsigned char *vf_buffer, int *vf_cr) { : unsigned long checksum; cs.b0 = vf_buffer[(*vf_cr)]; cs.b1 = vf_buffer[(*vf_cr) + 1]; cs.b2 = vf_buffer[(*vf_cr) + 2]; cs.b3 = vf_buffer[(*vf_cr) + 3]; (*vf_cr) += 4; checksum = (unsigned) (cs.b0 * 256 * 256 * 256 + cs.b1 * 256 * 256 + cs.b2 * 256 + cs.b3); :

if (checksum != 0 && font_checksum(k) != 0 && checksum != font_checksum(k)) vf_local_font_warning(f, k, "checksum mismatch", (int) checksum, (int) font_checksum(k)); if (ds != font_dsize(k)) vf_local_font_warning(f, k, "design size mismatch", ds, font_dsize(k));

It's not related to MKIV but to luatex, and it's a warning. It doesn't even matter to have --modules==all, this vf file in the standard standalone. Of course you must have the pfb/afm or otf or ttf file, as usual, otherwise the pdf is wrong

I was not sure, that why I began this thread. I started with latex example, then strip it down to plain tex one and after that came to context minimal example. AFAIU, context suite doesn't have "plain" pdftex and luatex formats, therefore I compile minimal plain tex example with TeX Live 2011, and luatex doesn't print such message though it reads vf file. The problem indeed may be it luatex and how it works with vf fonts, but as I said, I'm not sure (don't know very well either plain tex or context). P.S. Unrelated: seems I completely lost, I read web sources, the quoted extract from luatex source and I don't understand how this number '32212676346' can be presented as cs.b0 * 256 * 256 * 256 + cs.b1 * 256 * 256 + cs.b2 * 256 + cs.b3 If I don't lose my math skills the '32212676346' is 32212676346=7·256^{4}+128·256^{3}+6·256^{2}+110·256^{1}+250·256^{0} --- WBR, Vladimir Lomov -- If only God would give me some clear sign! Like making a large deposit in my name at a Swiss bank. -- Woody Allen, "Without Feathers"

Hello, ** luigi scarso [2012-02-01 15:04:07 +0100]: [...]

...

I was not sure, that why I began this thread. I started with latex example, then strip it down to plain tex one and after that came to context minimal example. it's also in pdftex: see function vf_def_font in src/texk/web2c/pdftexdir/pdftex.web

Will look in it.

...

AFAIU, context suite doesn't have "plain" pdftex and luatex formats, therefore I compile minimal plain tex example with TeX Live 2011, and luatex doesn't print such message though it reads vf file. Hm, have a look at tex/texmf-context/tex/generic/context/luatex

In my installation this directory doesn't have fmt files. In any case I will look in it.

...

The problem indeed may be it luatex and how it works with vf fonts, but as I said, I'm not sure (don't know very well either plain tex or context).

...

P.S. Unrelated: seems I completely lost, I read web sources, the quoted extract from luatex source and I don't understand how this number '32212676346' can be presented as

...

cs.b0 * 256 * 256 * 256 + cs.b1 * 256 * 256 + cs.b2 * 256 + cs.b3

...

If I don't lose my math skills the '32212676346' is 32212676346=7·256^{4}+128·256^{3}+6·256^{2}+110·256^{1}+250·256^{0} IIRC, 32212676346 should be the checksum stored inside the file, ie the font_checksum.

Yes, that's right, but I forgot that this is octal number, therefore 032212676346=3526065382=210·256^{3}+43·256^{2}+124·256^{1}+230·256^{0}, so 210,43,124,230 are written in file as D2, 2B, 7C, E6 (which I found in VF file :). --- WBR, Vladimir Lomov -- "The molars, I'm sure, will be all right, the molars can take care of themselves," the old man said, no longer to me. "But what will become of the bicuspids?" -- The Old Man and his Bridge

Am Wed, 1 Feb 2012 10:54:04 +0100 schrieb luigi scarso:

It's not related to MKIV but to luatex.

I think it is related to MKIV (or more precisely to the fontloader). It is a bit difficult to demonstrate it with context (how to you use luatex without mkiv there?) but latex (or plain) it is quite easy: The followings examples load pplr7t.vf. Both give the checksum mismatch message in the log if and only if the lua-fontloader luaotfload it also loaded: %LaTeX \documentclass{article} \usepackage{luaotfload} \begin{document} \font\test=pplr7t \test abc \end{document} %plain \input luaotfload.sty \font\test=pplr7t \test abc \bye So the code from luaotfload (which is based on the context fontloader code) changes a checksum (either in the vf or in the tfm-information) and so the check in the luatex engines fails. (Imho it is only a minor problem, vf-fonts are not much used with luatex). -- Ulrike Fischer

Hello, ** Hans Hagen [2012-02-01 15:22:58 +0100]:

On 1-2-2012 15:15, Ulrike Fischer wrote:

...

Am Wed, 1 Feb 2012 10:54:04 +0100 schrieb luigi scarso:

...

...

It's not related to MKIV but to luatex.

...

I think it is related to MKIV (or more precisely to the fontloader).

...

It is a bit difficult to demonstrate it with context (how to you use luatex without mkiv there?) but latex (or plain) it is quite easy:

...

The followings examples load pplr7t.vf. Both give the checksum mismatch message in the log if and only if the lua-fontloader luaotfload it also loaded:

...

%LaTeX \documentclass{article} \usepackage{luaotfload} \begin{document} \font\test=pplr7t \test abc \end{document}

...

%plain \input luaotfload.sty \font\test=pplr7t \test abc \bye

...

So the code from luaotfload (which is based on the context fontloader code) changes a checksum (either in the vf or in the tfm-information) and so the check in the luatex engines fails.

...

(Imho it is only a minor problem, vf-fonts are not much used with luatex).

Afaik nothing is done with a checksum. There is a checksum field in the loaded tfm but I don't think one has to be passed to luatex. Maybe one should be passed when a regular tfm file is used but even then, loading a vf file is independent.

So, it's best to just ignore that message.

Correct me if I'm wrong, but this how I understand TFM font and currect state of tex engines (actually pdftex, xetex and luatex): 1. to use any font in tex one need a TFM file (file name = fontname.TFM), that file actually contain informationabout font, not how exatly glyphs are constructed; 2. when [original] tex read a document file it searches for TFM and VF files, read them and write DVI file with information about that files; 3. after that user can send file to printer or publisher to print it on printer. As I understand the purpose of checksum was to be sure that publisher or printer would use exatly the same fonts as user. If user converts DVI file to PS/PDF one on his/she computer using dvips or dvipdfm* the checksum mostly useless, assuming files are not corrupted. Nowadays pdftex, xetex and luatex are widely used and most time users generate PDF files on the same computer they write documents, send PDF files which have they own mechanism to check font consistency. But still there are [plenty] DVI files around, as well as luatex engine might generate DVI file. The convertion to PS/PDF is performed by dvips/dvipdfm* programs, that's ok. But what about luatex with DVI output? My conclusion: 1. if PDF output is only interesting then it is Ok, ignore that message, because font information is already in PDF and PDF programs should deal with it; 2. if DVI output is concerned then luatex _must_ be consistent with pdftex (also can write DVI files), which, imho (don't check), takes care about both TFM and VF checksums. --- WBR, Vladimir Lomov -- The happiest time of a person's life is after his first divorce. -- J.K. Galbraith

On 2-2-2012 02:21, Vladimir Lomov wrote:

Correct me if I'm wrong, but this how I understand TFM font and currect state of tex engines (actually pdftex, xetex and luatex):

sort of wrong

1. to use any font in tex one need a TFM file (file name = fontname.TFM), that file actually contain informationabout font, not how exatly glyphs are constructed;

indeed, and when we forget about opentype etc only the tfm is needed

2. when [original] tex read a document file it searches for TFM and VF files, read them and write DVI file with information about that files;

no, the vf is only needed when the backend (which happens to be included in pdftex/luatex) is going to embed the glyph data from the font; only then it needs to know of a glyph is actually a virtual one however, in luatex, due to different internals, vf files are / can be read in earlier as the virtual font model is part of the front end so, it can be in an earlier stage that some mismatch can happen

3. after that user can send file to printer or publisher to print it on printer. As I understand the purpose of checksum was to be sure that publisher or printer would use exatly the same fonts as user. If user converts DVI file to PS/PDF one on his/she computer using dvips or dvipdfm* the checksum mostly useless, assuming files are not corrupted.

no, the checksum only is some safeguard that vf and tfm match if you go through dvi then dvips or dvipdfmx read the vf files

Nowadays pdftex, xetex and luatex are widely used and most time users generate PDF files on the same computer they write documents, send PDF files which have they own mechanism to check font consistency.

But still there are [plenty] DVI files around, as well as luatex engine might generate DVI file. The convertion to PS/PDF is performed by dvips/dvipdfm* programs, that's ok. But what about luatex with DVI output?

i never use luatex with dvi output

My conclusion: 1. if PDF output is only interesting then it is Ok, ignore that message, because font information is already in PDF and PDF programs should deal with it; 2. if DVI output is concerned then luatex _must_ be consistent with pdftex (also can write DVI files), which, imho (don't check), takes care about both TFM and VF checksums.

luatex probably is consistent although in most cases the way that it deals with fonts (read: the macro package deals with fonts) is different from the way it's done with pdftex/xetex if i had the time i'd probably run a few tests and see where the mismatch happens but it has a real low priority for me Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------

On Wed, Feb 01, 2012 at 03:47:32PM +0100, luigi scarso wrote:

On Wed, Feb 1, 2012 at 3:15 PM, Ulrike Fischer wrote:

...

Am Wed, 1 Feb 2012 10:54:04 +0100 schrieb luigi scarso:

...

It's not related to MKIV but to luatex.

I think it is related to MKIV (or more precisely to the fontloader). The message can be reproduced also with luatex --fmt=luatex-plain.fmt The font structure has a key checksum, but luatex doesn't use e by default is zero from vf from tfm used value type checksum yes yes no number default: 0 So maybe font_checksum is zero, and checksum is 32212676346 (i.e. the opposite that I wrote).

So now that is a question to Taco, if checksum is not used by luatex, why id there a difference when the font is loaded in the old way (i.e. no callbacks involved) and when it is loaded define_font callback? Regards, Khaled

Am Wed, 1 Feb 2012 17:23:09 +0200 schrieb Khaled Hosny:

I tried printing the tfm table we pass to the backed, and the checksum matches the one in the VF file, so this is a bit confusing

Well I actually don't know what test is actually done (and why exactly) but some remarks: 1. If I try to convert the binary "pplr7t.vf" to the readable vpl-file I need *two* tfm-files: pplr7t.tfm and pplr8r.tfm. 2. vftovp tells me during the conversion: "Check sum in VF file being replaced by TFM check sum" which probably means that the vpl-file doesn't contain the original checksum(s) of the vf-file. 3. The vpl file contains two checksums: (CHECKSUM O 25136566211) and a checksum in the mapfont entry: (MAPFONT D 0 (FONTNAME pplr8r) (FONTCHECKSUM O 36571141413) (FONTAT R 1.0) (FONTDSIZE R 10.0) ) So which of both is actually checked against which tfm checksum (and gives the mismatch message)? Btw: Two years ago I ran against a checksum mismatch message concerning the width of characters. In this case the culprit was a different calculation method for tfm and vf: http://tug.org/mailman/htdig/pdftex/2009-May/008035.html -- Ulrike Fischer

Basically it seems coherent with the luatex manual (ch. 7 Font structure): luatex read the txr.tfm file and set the checksum to the unsigned int 0 (luafont.w ,int font_from_lua(lua_State * L, int f)) (but see Note below) Then read the vf (dofont.w, do_vf(f) inside static int do_define_font(int f, const char *cnom, scaled s, int natural_dir)) do_vf(f) calculate the checksum of the vf file vua test_checksum() macro, and find that the vf file has checksum 210 43 124 230 =>(210*2^24)+(43*2^16)+(124*2^8)+230=3526065382= oct 32212676346 vs 128 0 0 0 and hence the warning of mismatch. Note: The last sequence 128 0 0 0 (i.e. 0) it is due to i = numeric_field(L, "checksum", 0); set_font_checksum(f, (unsigned) i); where static int numeric_field(lua_State * L, const char *name, int dflt) { int i = dflt; lua_pushstring(L, name); lua_rawget(L, -2); if (lua_isnumber(L, -1)) { i = lua_roundnumber(L, -1); } lua_pop(L, 1); return i; } and #define lua_roundnumber(a,b) (int)floor((double)lua_tonumber(L,-1)+0.5) I think one can avoid the warning if set the checksum field to the correct value by the define_font callback on the lua side. -- luigi