On Fri, 02 Nov 2001 04:09:35 +0900, "Stephen J. Turnbull" said: Given that fontsets exist to solve an actual problem (that you may not have one font that covers the charspace, but can use several related fonts to cover), what do you propose to replace the fontset support that you wish to get rid of? -- Valdis Kletnieks Operating Systems Analyst Virginia Tech

On Fri, 02 Nov 2001 04:38:50 +0900, "Stephen J. Turnbull" said: Superior it may be. However, checking the FAQ that came with XEmacs: Q3.2.2: How do I set the text, menu and modeline fonts? ------------------------------------------------------- Note that you should use `Emacs.' and not `Emacs*' when setting face values. In `.Xdefaults': Emacs.default.attributeFont: -*-*-medium-r-*-*-*-120-*-*-m-*-*-* Emacs*menubar*font: fixed Emacs.modeline.attributeFont: fixed This is confusing because modeline is a face, and can be found listed with all faces in the current mode by using `M-x set-face-font (enter) ?'. It uses the face specification of `attributeFont', while menubar is a normal X thing that uses the specification `font'. With Motif it may be necessary to use `fontList' instead of `font'. Hmm... that was helpful. Really. Doesn't explain how to use a face instead of a fontset. There's nothing about faces in the "Internationalization" section of the FAQ. There's nothing about setting faces for this in the sample.emacs OK.. Let's go look around in the menubar. "Options" has "Edit faces", go poking around there for a while, and finally find 'default' face in the 615-line buffer. Right mouse on the little arrow, and *aha*, we have 'Font Family'. Hmm... what does it want? I can *guess* that it wants 'Size' as either pixels or points - but it's unclear which. The other attributes.. Hmm.. pixmap it probably wants a pathname. A bunch of toggles. Foreground/Background, I'm *guessing* it wants a color specifier, not clear if it wants an #rrggbb value, or a (1.0,1.0,1.0) style, or a name from the rgb.txt file, or whether it's smart enough to guess what I mean. But "Font Family"? Say I want Helvetica - can I specify 'adobe-helvetica', or do I have to hope it doesn't guess helvetica from the alias, cronyx, or urw foundries? Or does it want 'adobe helvetica'? Hmm.. that might be fun if I wanted 'new times roman' that has blanks in the XLFD.... (As an aside - I did 'control-h ? ?' and here's what it *said*: C Customize Emacs options. C-d How to obtain XEmacs. e Information about the most recent error. F Local copy of the XEmacs FAQ. Hmm... so C-h C gets us help on how to *use* this Customize? Nope, it *invokes* customize. Blech. The *other* choices near it in the list produce *info* - control-h C-d doesn't download XEmacs, it tells you how to do so). So the score is now Customize 2, user 0. OK.. 'C-h a face', and start digging. Not much stuff that has the little asterist that says it's user-callable. Going and looking at the docstring for set-face-font points you to make-font-specifier. That says: Compiled Lisp function, (loaded from "/home/valdis/src/xemacs-21.2/lisp/objects.elc"): arguments: (spec-list) Return a new `font' specifier object with the given specification list. SPEC-LIST can be a list of specifications (each of which is a cons of a locale and a list of instantiators), a single instantiator, or a list of instantiators. See `make-specifier' for more information about specifiers. Valid instantiators for font specifiers are: -- a string naming a font (e.g. under X this might be "-*-courier-medium-r-*-*-*-140-*-*-*-*-iso8859-*" for a 14-point upright medium-weight Courier font) -- a font instance (use that instance directly if the device matches, or use the string that generated it) -- a vector of no elements (only on TTY's; this means to set no font at all, thus using the "natural" font of the terminal's text) -- a vector of one element (a face to inherit from) Now... how is a poor novice user who just *knows* "this fontset works just fine for me everyplace else" supposed to figure out how to specify a face without using a fontset? OK.. Maybe the answer is in the User's Guide someplace, I'll have to investigate that once I figure out why RedHat's texinfo-4.0c-1 does this: make[1]: Entering directory `/home/valdis/src/xemacs-21.2/man' makeinfo -P xemacs -o ../info/xemacs.info xemacs/xemacs.texi xemacs/programs.texi:1036: First argument to cross-reference may not be empty. makeinfo: Removing output file `../info/xemacs.info' due to errors; use --force to preserve. make[1]: *** [../info/xemacs.info] Error 2 /Valdis (who just spent 45 minutes *trying* to implement the superior solution).

On Fri, 02 Nov 2001 11:04:53 +0900, "Stephen J. Turnbull" said: OH.. OK.. Having a "face" is OK by me, and if the UI allows me to say "Use this FontSet" for the face, and things get displayed on the monitor the way they should be, I'm happy. I get paid to worry about the internals of enough OTHER things that I've intentionally *tried* to avoid learning more about the innards of XEmacs than what I need to answer the question "What do I need in the init.el to make it do THIS?" ;) In my shop, that's always called "In my copious spare time..." ;) No, actually, now that I know that my failure to figure it out was not my stupidity, but the fact you can't GET there from here.. ;) I spent quite some time getting exmh to display stuff right, even though I'm only truly fluent in one language. I figure that if it gets displayed the way the sender intended, then any lack of understanding is my fault, not the computer's... But see my comment above about free time.. ;) Man, I'm on a roll here ;) I seem to be unable to go anywhere NEAR a piece of code without leaving a trail of destruction behind me - I spent most of the last 48 hours finally (I think) tracking down a particularly nasty intermittent Sendmail bug - before finding a < that should probably be <=, I managed to find at least 3 other things broken in the code.. At least I'm not as bad as my co-worker Randy. He once plugged a hot-swap disk into a Dec 2100, and got a conflicting address (neat trick, since the addresses were assigned by the cabling based on the media bay). Tried another bay, it conflicted with a *differnt* disk. Further experimentation found that it was *not* a stuck address bit. The drive decided if the backplane thought its address was N, that it would answer to ((N+3) mod 7). We still dont know why.... /Valdis