Hello, Alan Cox, le Mon 20 Apr 2009 11:29:37 +0100, a écrit : > On Sat, 18 Apr 2009 23:06:38 +0200 > Samuel Thibault <samuel.thibault@xxxxxxxxxxxx> wrote: > > A chinese user reported that some characters don't show up fine on > > its console. This happens to be because character 128+27 is handled > > specially by the linux Console. I believe the same issues can happen > > with other characters sets. Since linux' terminfo makes applications > > use \E[ anyway, would it be ok to add an option to disable the 128+27 > > handling? > > I'm reluctant to do so because its extra complication we don't need (at > least nobody has needed for over fifteen years), This has been unseen for years because there haven't been chinese blind users reporting it sofar, there's only one who has showed up relatively recently on the brltty mailing list. Let me explain a bit more: sighted chinese users wouldn't use the plain linux console, as the 512-glyphs limitation doesn't permit to display kanjis, so they use things like zhcon or just Xorg, which do not use vt.c's code. Blind chinese users, however, need their screen reader to be able to know what's on the screen. What zhcon displays on the screen is however not available to the reader as it's just an matrix internal to zhcon, so these user have to either use Xorg (which is far from easy), or just use the plain console in 1:1 mapping mode and not care about the glyphs that get rendered by the video board, while the screen reader, however, can in that case correctly interpret BIG5. > it's bound to confuse some other application, Well, concerning _displaying_ text on the screen, applications don't care, only the kernel does. And as I raised, because of the linux terminfo entry, applications are not supposed to emit CSI through 128+\E, but \E[. Concerning typing text, yes that might confuse layers that would interpret 128+\E as being CSI. These would need the same kind of option (which could be automatically enabled in BIG5 locales). The fact that zh_TW BIG5 is in my /etc/locale.gen actually tells me that for now we really have a bug that chinese people just haven't taken the time to raise. > and given almost everyone has already gone UTF-8 it's really debating > a corner case only relevant to the past anyway. I would tend to agree, except that in the case at stake (a blind chinese user), UTF-8 just can not be used, because /dev/vcsa does not provide the unicode characters that the linux kernel tries to show on the screen when the font doesn't have enough glyphs to actually show it. That's not a problem for most languages which can cope with the 512 glyphs limit, but it's one for chinese. On the other hand, using the 8bit 1:1 mapping just works, except the CSI issue. Yes, the /dev/vcsa interface should ideally be rethought, but it's more than just defining a new interface, since the kernel itself uses the VGA representation... I haven't been able to take the time to fix that yet of course... Samuel -- To unsubscribe from this list: send the line "unsubscribe linux-console" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
