Turned out that our ARM gcc4 build used short-enum's, which made
our build fail (specifically, src/kits/tracker/ViewState.cpp:122).
Thanks to Landon Fuller and Oliver Tappe for helping me figure
the fix out!
These are part of the C++ standard, I don't know how they went missing
in our version.
http://www.cplusplus.com/reference/std/complex/
They are needed to build MathGL library, and likely other stuff.
* O_BINARY and O_TEXT were removed from Haiku way back in hrev38962
* This patch removes the fix that used to unset those flags
* Now gcc/intl/loadmsgcat.c is back to the GCC 4.6.3 stock file
Thanks diger for noticing the previously needed fix was still in place unintentionally
* -lssp_nonshared being passed at link time was resulting in multiple definition errors
* Apps such as Perl and CUPS would fail to build due to this
* This should alleviate the problem when -fstack-protector or -fstack-protector-all are used
Thanks to diger for letting me know of this issue and making me aware of a fix
* This had the side effect of doing the opposite of the intended goal
* Now the GCC package is smaller, as before there was a lot of static linking happening
Thanks to diver for pointing out the size change of the GCC package, as it helped me find this issue.
* This changes the optimization level from -O3 to -O
* This will prevent segfaults when creating jam via the Makefile on FreeBSD
* Benchmarking was done, and there seems to be no negative performance impact
* Another option would be -O3 -fno-strict-aliasing, but this has no noticeable benefit over -O
* This update resolves the segfault/ICE in gcc/c-parser.c while building natively
* gcc/gengtype.c was already updated to GCC 4.6.3 version previously in btrev43031, so no change there
* gcc/libiberty/make-relative-prefix is the stock 4.6.3 version now, as the workaround for Haiku is no longer needed
* This update is otherwise the GCC patches for upgrading 4.6.2 to 4.6.3 core/g++ applied to our sources
* A build failure occurs as TARGET_64BIT is no longer a compile-time
constant in GCC.
* Made some changes so that generated code matches the i386 Haiku target when
generating 32-bit code, and follows the x86_64 ABI for 64-bit code.
* Added a comment stating that DEFAULT_PCC_STRUCT_RETURN needs to be defined
to follow the ABI for 32-bit code, it is ignored for 64-bit.
* Remove definition of __x86_64__, already defined by GCC's i386-c.c.
* Remove -z max-page-size from LINK_SPEC as it shouldn't be necessary: the
default page size is 2MB but this won't result in large binaries, it just
changes the alignment of the sections and would make it easier to map
binaries using large pages in the future.
Signed-off-by: Ingo Weinhold <ingo_weinhold@gmx.de>
* arm: ASM_COMMENT_START is defined in arm/aout.h.
* i386 and x86_64: define ASM_COMMENT_START.
* others: ASM_COMMENT_START is defined in ${arch}/${arch}.h.
If you take a release tarball (which has pregenerated syslex and sysinfo files
in it), apply some patches which touch syslex.l, and then build the result out
of tree, it will fail. This is because syslex.l uses sysinfo.h, but the
sysinfo.y file wasn't updated and so it wasn't regenerated (the files are found
in the $srcdir), and the build rule for syslex.c does not use -I$(srcdir) when
it finds a local file. Simple fix below.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
