[HarfBuzz] icu-le-hb

[Steven R. Loomis]

On 15/11/13 11:06, Behdad Esfahbod wrote:
> On 13-11-15 01:49 PM, Steven R. Loomis wrote:
>> On 15/11/13 10:26, Behdad Esfahbod wrote:
>>> On 13-11-15 01:15 PM, Dejan Kovacevic wrote:
>>> Doesn't really matter as far as ICU goes - I don't think it would
>>> ever be diffed against ICU's code itself, so it's better as Behdad
>>> said to keep it closer to the HB code. 
>> ...
>> back to the technical side -- does it matter that HB sometimes
>> depends on ICU? That seems to be the only obstacle I see to having a
>> "--use-harfbuzz" type switch on the ICU LE. Let me know what you think. 
> Good point.  We already have that problem with FreeType...  harfbuzz-icu is a
> separate library, but yeah, makes it hard to build.  For now I think having
> icu-le-hb as a separate module is good enough.  We should document, for those
> who rather build a combined version, how they can do that.

Yeah, it's sort of six of one = half a dozen of the other. I thought
briefly that it could help users of ICU to be able to just add
'--use-harfbuzz' and it would just find HB on their system.  Then the
switch is in ICU, instead of their code having to have a  "use ICU but
really use harfbuzz'. 

On the other hand, with the pkg-config layout we discussed, I think we
ended up having 'icu-le-hb' as a module name, so then it should just be
a matter of choosing  'icu-le' vs 'icu-le-hb' at link/include time.

Let's go with path of least resistance here. So I propose, that you look
at merging/replacing the wrapper with Dejan's, managing whatever
formalities are needed.

Also FWIW I opened ticket here,
http://bugs.icu-project.org/trac/ticket/10530  so that ICU makes the
policy more clearly known.

>From a testing point of view, I need to run the icu letest regression
test through the wrapper.. that would be a good test. Maybe first
separate that test out.  Anyways, the data is here -
http://source.icu-project.org/repos/icu/icu/trunk/source/test/testdata/letest.xml
(it's generated).

-s