minutes of ESC face-to-face ...

[Stephan Bergmann]

On 09/25/2013 02:30 AM, Michael Meeks wrote:
> * release builds (Norbert)
>      + propose continue doing 10.6 for 32bit, and 10.8 for 64bit
>      + more and more SDK pieces are harder and harder to get to
>        work on old releases
> AI:     + switch to 10.8 baseline for 64bit (Norbert)
>      + ultimately, gerrit will do both with DavidO's help

One consequence of a Mac OS X 10.8 based build that comes to mind is 
that Clang does not support GCC's -fno-enforce-eh-specs.  So since a 
10.8 based build would necessarily(?) use a true Clang, we would be back 
with a platform where even --disable-dbgutil builds could run into 
std::unexpected aborts.  (Those exception violations appear to be rare 
in practice, and are bugs that need to be fixed anyway, so this should 
not be too big an issue; just want to mention it.)

(According to e.g. 
<http://dev-builds.libreoffice.org/daily/libreoffice-4-1/MacOSX-Intel@27-OSX_10.7.0-gcc_4.2.1_llvm/2013-10-14_10.42.26/libreoffice-4-1~2013-10-14_10.42.26_build_info.txt>

> checking whether /usr/local/bin/ccache g++-4.2 -m32 -mmacosx-version-min=10.6 -isysroot /Developer/SDKs/MacOSX10.6.sdk supports -fno-enforce-eh-specs... yes

the 10.6 based builds apparently use an (Apple-variant) GCC that still 
supports -fno-enforce-eh-specs.)

Stephan