Jerry> I had a few free minutes this afternoon (okay, so I *took* Jerry> a few free minutes this afternoon), and poked around in the Jerry> coding system stuff to see what's up with the LFD/RET Jerry> problems on TTYs. I found that TTYs get the 'undecided Jerry> coding system, and that the 'undecided coding system has Jerry> the 'do-eol property, which triggers EOL autodetection. OK. I don't really understand the new detectors, but I suspect this is a bug in them. There are dire warnings about attempting to detect on non-seekable streams. Since when are TTYs seekable? Jerry> Changing from 'undecided to 'undecided-unix (which does not Jerry> have the 'do-eol property) seems to be the right way to go. I dunno. Do you know for a fact that there are no TTYs out there that send CRLF? Jerry> It looks like this would be a problem on Windows machines Jerry> with MULE as well. Can somebody confirm or deny that? Erm, I can't get XEmacs to run in a TTY on a Windows box. Jerry> The patch below seems to fix the problem for me, but as I Jerry> don't know this code terribly well, I'd appreciate a few Jerry> more eyes looking at it. I think that the right thing to do is to fix the detectors, but this is not terribly harmful (the worst that could happen is stair-stepping, I suspect). However, all appearances to the contrary, "raw-text" does not mean "binary", it means "autodetect EOL." So if setting the EOL to -unix makes sense, both of those should be "binary". (I would actually go so far as to put all of the keyboard "binary". I'm pretty sure autodetection fails in a uxterm.) Bottom line, I'd like to hear from Ben, but if Steve Y makes an executive decision to put this in it's OK by me. Jerry's patch for Ben's benefit: ------------------------------------------------------------------------ lisp/ChangeLog addition: 2003-05-08 Jerry James <james(a)xemacs.org&gt; * code-init.el (coding-system-variable-default-value-table): Use undecided-unix for keyboards to avoid EOL autodetection. * code-init.el (coding-system-variable-default-value): Update the comment to reflect the changed table. xemacs-21.5 source patch: Diff command: cvs -q diff -uN Files affected: lisp/code-init.el Index: lisp/code-init.el =================================================================== RCS file: /pack/xemacscvs/XEmacs/xemacs/lisp/code-init.el,v retrieving revision 1.5 diff -d -u -u -r1.5 code-init.el --- lisp/code-init.el 2003/02/21 06:56:47 1.5 +++ lisp/code-init.el 2003/05/08 22:15:35 ＠＠ -79,7 +79,7 ＠＠ (native binary binary binary raw-text-dos mswindows-multibyte-system-default-dos) (keyboard - binary raw-text undecided raw-text undecided) + binary raw-text undecided-unix raw-text undecided-unix) ;; the `terminal' coding system is used for output to stderr. such ;; streams do automatic lf->crlf encoding in the C library, so we need ;; to not do the same translations ourselves. ＠＠ -158,16 +158,16 ＠＠ The table of default values looks like this: (see below for abbreviations) - Unix Unix+EOL Unix+Mule MSW MSW+Mule ------------------------------------------------------------------------------- -bfcs-for-read binary raw-text undecided raw-text undecided -default bfcs binary binary iso-2022-8 raw-text-dos MSW-MB-dos -native binary binary binary raw-text-dos MSW-MB-SD-dos -keyboard binary raw-text undecided raw-text undecided -terminal binary binary binary binary MSW-MB-unix -process-read binary raw-text undecided raw-text undecided -process-write binary binary binary raw-text MSW-MB-SD -no-conv-cs binary raw-text raw-text raw-text MSW-MB + Unix Unix+EOL Unix+Mule MSW MSW+Mule +----------------------------------------------------------------------------- +bfcs-for-read binary raw-text undecided raw-text undecided +default bfcs binary binary iso-2022-8 raw-text-dos MSW-MB-dos +native binary binary binary raw-text-dos MSW-MB-SD-dos +keyboard binary raw-text undecided-unix raw-text undecided-unix +terminal binary binary binary binary MSW-MB-unix +process-read binary raw-text undecided raw-text undecided +process-write binary binary binary raw-text MSW-MB-SD +no-conv-cs binary raw-text raw-text raw-text MSW-MB VAR can be one of: (abbreviations in parens) -- Jerry James http://www.ittc.ku.edu/~james/ -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Ask not how you can "do" free software business; ask what your business can "do for" free software.

|--==> "JJ" == Jerry James <james(a)xemacs.org&gt; writes: JJ> A couple of hours after I sent this patch in, a tornado touched down JJ> about 1 block from my house. My place is fine; there isn't so much as a JJ> roof tile missing. A block away, an apartment complex was more or less JJ> destroyed. The roofs are heavily damaged or missing, and the entire JJ> second floor of one building is just gone. There were no deaths or JJ> serious injuries, thanks to the excellent advance warnings we received. JJ> Even so, does anybody have a job opening in California? I grew up with JJ> earthquakes; I can handle them a lot better. :-) Australia doesn't have earthquakes or tornados, should I watch the positions vacant section of the newspaper for you? :-P JJ> "Stephen J. Turnbull" <stephen(a)xemacs.org&gt; wrote: Decision's made. Ben has been ignoring this bug for more than a year now, countless people have reported the bug, I personally have asked Ben countless times to fix it. Commit the patch, Jerry. -- |---<Steve Youngs>---------------<GnuPG KeyID: 10D5C9C5>---| | XEmacs - The only _______ you'll ever need. | | Fill in the blank, yes, it's THAT good! | |------------------------------------&lt;youngs(a)xemacs.org&gt;---|

Jerry> I grew up with earthquakes; I can handle them a lot Jerry> better. :-) Hey, I can give earthquakes, and active volcanos, too! Jerry> Yes, I saw those warnings. But then the 'keyboard streams Jerry> are explicitly mentioned all over the place, too. It's Jerry> confusing. Around here, we spell that "Winging code". ;-) I happen to like it once I understand it, but that first step is a doozy. Jerry> Nope. I have no idea. But even if there are, how are you Jerry> going to figure it out? You assume that if in the first few lines you always see CRLF, it's the terminal that's doing it. Surely no user would. However, that requires an online detector, and ours require seek, so ... you don't. Jerry> I don't know that the detectors need fixing. By "fix", I meant "like you fix a cat." }:^} Jerry> Hmmmm... the second column is Unix+EOL, which arguably Jerry> should not be used for keyboards at all. The columns are determined by what capabilities are compiled in. The entries decide what capabilities are actually used. Jerry> Even so, I'd feel better leaving it "raw-text". raw-text = "binary except do evil things with newlines". Ie, raw-text-unix == binary. Jerry> I think you are right about the MSW column, though. That Jerry> should be "binary". But isn't "undecided-unix" right for Jerry> the MULE variants, or do coding systems have nothing to do Jerry> with input methods? (Hopefully that question doesn't Jerry> betray hopeless ignorance on my part!) If the comments about not doing detection on non-seekable streams mean anything here, undecided-unix is (in effect) the same as binary. I say, let's make this explicit, until somebody can authoritatively say what's going on. Ben being silent, and me being some weeks away from enlightment, I would guess ... any other volunteers? -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Ask not how you can "do" free software business; ask what your business can "do for" free software.