Re: Thai support in framebuffer d-i
OK, I'm convinced.
Sorry if I seemed negative about this, but I really thought there could be
no way you were seeing this with a regular image as we're supposed to be
only including characters needed by supported languages in the unifont
included in the installer.
However, it turns out that the build system worked slightly different than
I thought, which meant glyphs that shouldn't be included actually are
being included. But I've fixed that today (resulting in e.g. a 300kB size
reduction for a regular amd64 netboot image).
I also would have thought enabling Thai for the regular installer could
have been done much sooner as I don't see what would have changed recently
that "suddenly" makes this possible now.
You can actually quite easily run an installation in Thai using an existing
image, even using a Lenny image:
- boot the installer with BOOT_DEBUG=3 (you'll get a debug shell)
- cd /usr/share/localechooser
- zcat localechooser.data.gz > localechooser.data
- use nano to edit that file and change "4" to "3" in the line for Thai
- use nano to edit /usr/bin/localechooser
- remove ".gz" in the line that sets LANGUAGELISTDATA
- change "zcat" to "cat" in the line that has 'zcat $LANGUAGELISTDATA'
(using ^W to search for LISTDATA helps)
- type 'exit' twice to start the installation
- choose Thai from in the language selection dialog
You should not see any missing characters early on, but you may after the
stage where additional components are loaded.
The next question is: do we actually want to enable Thai in the regular
installer. IMO the answer is yes, especially as it is effectively already
supported and will not result in a size increase of images.
Also, I see no reason why e.g. Vietnamese should be supported and Thai not.
I'm going to ask Christian to actually manage the change. AFAIK there's
three things that need to be done:
- we'll need a proper "needed-characters" file for Thai from you
- switching the level in localechooser
- removing th from GI_LANGS in build/config/common
On Thursday 11 February 2010, Theppitak Karoonboonyanan wrote:
> I've classified the missing glyphs in di-missing.txt.
> - "In-use" means it's currently used in d-i translation; so it's
> - "Potential" means it's not found in d-i translation, but there is high
> potential to be used in other relevant packages (xorg, exim4, debconf,
> - "Rare" means it's rarely used, and can be safely omitted.
Thanks. For the "needed-characters" file we'll need a file containing *all*
common characters, including ones already included. See files in  for
examples for other languages and  for info.
> For the missing Thai digits and punctuation marks, I'm quite sure
> they are not used in the translation.
> For the screenshot, I've chosen a bitmap font that is closest to the one
> used in the directfb d-i. It's an 8-bit TIS-620 font aliased "thai8x16"
> from xfonts-thai-manop package. And I took it on xiterm+thai (from
> xiterm+thai package) by converting it to TIS-620 for the display.
I've quickly built an image using that file so the missing characters would
There was still one missing character: 4th line, 5th from the end. But it
was there when I used your testmsg-di file during the build. This shows
the importance of a complete "needed-characters" file.