[HarfBuzz] [kakuto at fuk.kindai.ac.jp: new XeTeX]

[Behdad Esfahbod]

On 13-01-04 05:43 AM, Peter Breitenlohner wrote:
> On Fri, 4 Jan 2013, Behdad Esfahbod wrote:
> 
>> I applied a change that should address the second part of your patch (the
>> const's). ...
> 
> Hi Behad,
> 
> as for the first part of Akira's patch I found this:
> 
> in libharfbuzz.a for linux there are (nm -C libharfbuzz.a | grep ::Cmp::)
>   int OT::SortedArrayOf<OT::RangeRecord>::search<uns...
>   int OT::SortedArrayOf<OT::IntType<unsigned short, 2u> >::search<uns...
> in both hb-ot-layout.o and hb-ot-shape-complex-arabic.o as global weak
> symbols (W), whereas in libharfbuzz.a from MinGW32 gcc-4.6.3 they are
> non-weak (T), however, the GNU linker seems to be able to handle this
> situation, but it seems the MSVC linker produces an error.
> 
> I think one has to live with the fact that there are systems without weak
> symbols, and duplicate global non-weak symbols are certainly a bad idea
> (even if GNU ld seems to be able to handle them).

Thanks.  But, how do people deal with template instantiations on MSVC then?
What's happening is that the same template is being instantiated using the
same types from two different compilation units.  This is AFAIU the most
common pattern of using templates, isn't it?


> ============================
> 
> When building harfbuzz-0.9.10 (independent of TeX Live and XeTeX) I had to
> run ./autogen.sh (with Autoconf 2.69 and Automake 1.13 or 1.13.1) and got
> several warnings.
> 
> (1) You could add AM_PROG_AR to configure.ac to remove some of them.
> This requires Automake 1.11.2 or better.
> 
> (2) And you could replace
>     valgrind_verbose = $(valgrind_verbose_$(V))
>     valgrind_verbose_ = $(valgrind_verbose_$(AM_DEFAULT_VERBOSITY))
> in test/api/Makefile.am by
>     valgrind_verbose = $(valgrind_verbose_ at AM_V@)
>     valgrind_verbose_ = $(valgrind_verbose_ at AM_DEFAULT_V@)
> to remove the rest of them.
> This requires Automake 1.11.3 or better.

Thanks.  I'll look into this.

behdad


> Regards
> Peter Breitenlohner <peb at mppmu.mpg.de>
> 

-- 
behdad
http://behdad.org/