On 03/05/2016 05:38 PM, Pranav Ganorkar wrote:
When trying to build sfx2 module with debugging symbols , I get the
following build error:
make sfx2.clean && make sfx2 debug=t
I fear that "selective debug" thing is not really working fully.
The problem is that recent GCC has a bug
(<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66460> "ICE using
__func__ in constexpr function") that causes internal compiler errors in
constexpr functions that contain calls to assert (and where NDEBUG is
not defined, so that the calls to assert actually expand to code).
That's why LO's configure.ac tries to detect whether marking functions
that contain calls to assert as "constexpr" (via the expansion of LO's
SAL_CONSTEXPR macro) is OK (see HAVE_CXX14_CONSTEXPR in
config_host/config_global.h).
If you build LO with a GCC with above bug, using constexpr is still fine
as long as calls to assert don't actually expand to any real code (i.e.,
in the implicit --disable-debug case). Hence, autogen.sh originally
determined that HAVE_CXX14_CONSTEXPR is 1 for your environment. But
when you call "make ... debug=t", autogen.sh does not get re-run, so the
problematic functions (in include/o3tl/typed_flags_set.hxx) still get
marked as "constexpr," even though their calls to assert now expand to
code that triggers the GCC bug.
Probably best to disable HAVE_CXX14_CONSTEXPR regardless of whether your
configure switches would cause NDEBUG to be defined, so that later
builds with "make ... debug=t" do not run into this problem. Did that
now with
<https://cgit.freedesktop.org/libreoffice/core/commit/?id=4b8c29015d7c70121a161da0f3a85fa4ea0c5987>
"Make detection of HAVE_CXX14_CONSTEXPR work with 'selective
debugging.'" (Do a full top-level "make" once after pulling that, and
afterwards "make sfx2.clean && make sfx2 debug=t" should work for you.)
_______________________________________________
LibreOffice mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
