Hi Pali, Pali Rohar wrote on Sun, May 27, 2018 at 11:52:44PM +0200:
> Now I looked deeply at man -Tps output and basically \- sequence is > written as character 0xAD (\255 in octet) into output postscript file. > Therefore it is SOFT HYPHEN (U+00AD), No, that is not a "soft hyphen". Glyph numbers in fonts used for PostScript output have nothing to do with Unicode code points. Look at the file font/devps/TR for examples: PS name TR# Unicode ------- --- ------- asciicircum 0x00 U+005E asciitilde 0x01 U+007E Scaron 0x02 U+0053 U+030C Zcaron 0x03 U+005A U+030C scaron 0x04 U+0073 U+030C zcaron 0x05 U+007A U+030C Ydieresis 0x06 U+0059 U+0308 trademark 0x07 U+2122 quotesingle 0x08 U+0027 Euro 0x09 U+20AC hyphen 0x2d U+2010 circumflex 0x5e U+02C6 quoteleft 0x60 U+2018 tilde 0x7e U+02DC bullet 0x83 U+2022 florin 0x84 U+0192 minus 0xad U+2212 and so on and so forth, it's completely different all over the place. > so it is incorrect for command line switch. It is not incorrect. The TR font does not contain a glyph for hyphen-minus, so plain minus is used as a fallback. > I looked deeply at DVI output and basically \- is printed in DVI as > glyph with index 0 from font cmsy10. Looking at cmsy10.pfb Type1 font > file and there is "dup 0 /minus put" which means that this character is > mathematical minus with Unicode code point U+2212. That could well be the same effect as for PostScript. I don't have the slightest idea how the dvi file format works (even though i used LaTeX a lot in the past). Is the wish "i want to print hyphen-minus" reasonable for dvi? It might be just as meaningless as for PostScript with the default TR font. > So the result is that "preferred" way for writing command line switches > in manpage via \- sequence is broken for -Tps and -Tdvi. Sure, for PostScript, that breakage is expected and can't be fixed. No idea with respect to dvi. > Is somebody going to fix PS That's impossible. > and DVI output to be "compatible" with "preferred" way how > to write hyphen-minus in manpages? I don't know about dvi. HTML could probably be fixed if somebody cared enough to track down the root cause of the bug, though i'm not completely sure how that works internally, so maybe i'm too optimistic. Yours, Ingo
