Ah, I was reading the following and came across something that is an argument against version scripts.
https://gcc.gnu.org/wiki/Visibility In particular: "Furthermore, using linker version scripts doesn't permit GCC to better optimise the code." On Wed, Jan 31, 2018 at 3:41 PM, Tom Kacvinsky <tkacv...@gmail.com> wrote: > Just let me know if you want me to proceed. I am not going to spend the > time on this if the consensus (sans me) is that we don't need/want it. > > On Wed, Jan 31, 2018 at 3:22 AM, Werner LEMBERG <w...@gnu.org> wrote: > >> >> > I also hated the frontend of the symbol versions which requires GCC >> > assembler trickery. >> > >> > https://gcc.gnu.org/wiki/SymbolVersioning >> > >> > This is a can of warms, which I do not want to open. >> >> I guess we don't need this. >> >> > I have started this when I saw that libtool wraps the flat file of >> > symbols to export into an anonymous version script. That seemed >> > weird. Right now I would actually prefer >> > __attribute__((__visibility__("default"))) to any list, as gcc and >> > Vincent suggest >> > >> > https://www.gnu.org/software/gnulib/manual/html_node/Export >> ed-Symbols-of-Shared-Libraries.html >> > http://lists.nongnu.org/archive/html/freetype-devel/2018- >> 01/msg00067.html >> >> I don't object, and it could be integrated into the `FT_EXPORT' macro, >> AFAICS. However, this doesn't give versioned symbols. IIRC, Drepper >> recommends to use both approaches: the `__visibility__' attribute >> improves code generation of the compiler, and the map file provides >> symbol versioning. >> >> >> Werner >> > >
_______________________________________________ Freetype-devel mailing list Freetype-devel@nongnu.org https://lists.nongnu.org/mailman/listinfo/freetype-devel
