bug#4895: 23.1; Incorrect font selected

[Chong Yidong <cyd@gnu.org>]

Juri Linkov <juri@jurta.org> writes:

>>> Thanks for the explanation. Do you think it's still valid not to
>>> prefer unicode fonts to fonts with specific registries? I believe the
>>> unicode is dominating by now, at least for european languages.
>>
>> I'm not sure.  At least, a while ago, bitmap fonts of legacy
>> registries (e.g. iso8859-X, koi8, etc) are better than
>> unicode bitmap fonts.  I think using cp1251 fonts for
>> cp1251-encoded file is a good default even now.  First of
>> all, if unicode is dominating by now, the chance of reading
>> cp1251 files should be rare.
>
> Unfortunately, cp1251 is still common in the documents created on
> Windows.  Opening such documents in Emacs on GNU/Linux often results
> in a very ugly look, especially when cp1251 text is surrounded
> by Latin or Unicode characters.  Unicode fonts are well developed and 
> maintained unlike legacy cp1251 fonts that are less relevant nowadays.

Presumably, unicode fonts should have even better coverage now than
three years ago, when this discussion first took place.  Any objections
to making the switch?

Also, where in the code does this

>> when you read a file with cp1251, the charset property
>> `cp1251' is put on the buffer text.  Thus the font-selector
>> prefers a font whose registry is "microsoft-cp1251".

occur?  I took a quick look at the documentation, e.g. the docstring of
new-fontset, and couldn't find the explanation of how the font-selector
chooses the font to use from the list of font specifications.  If the
list of font specifications is not given in order of priority, that
should be clearly explained somewhere.