While here:
- adjust python scripts shebang lines
- also byte-compile the .py files in the lldb subpackage. One file
needs a brute-force conversion using 2to3
Tested in a bulk by ajacoutot@, prompted by and ok kn@
The installed headers don't need it, except for a testsuite header not
used by the ecosystem.
Also drop two outdated PLIST lines from back when llvm tried to install
libgtest.a
Spotted while discussing the deps of the various subpackages with kn@
Tested in a bulk by ajacoutot@, ok kn@
- don't install six.py, turns out there's a cmake config knob for that
- drop misplaced, @commented out header
- add @so and @static-lib markers
Don't use DEBUG_PACKAGES for now, as the size increase of ${WRKBUILD} is
huge. I tried to use -gline-tables-only (I think?) as advised by kmos@
but the size of the resulting non-debug packages increased. To be
revisited...
- Add retguard for octeon/mips64.
- Move the hashed __retguard_* symbols into individual sections and mark
them as COMDATs so that the linker can individually discard them, instead
of just ignoring duplicate symbols but keep the (duplicate) space.
from Brad
- Make clang emit the proper path to our libcompiler_rt.a when asked to.
- Sort relative relocations (and relocations against the same symbol) by offset
to optimize the cache and UVM faulting behavior
from Brad
Passing -DBacktrace_LIBRARY='' doesn't work since the flag has been
renamed to Backtrace_LIBRARIES and isn't enough anyway to prevent the
llvm cmake files to *add a dep on all headers* in
${Backtrace_INCLUDE_DIRS}, ie /usr/local/include.
... which of course leads to builds blowing up whenever dpb(1) junking
kicks in.
Bulk builders: sorry it took me so long to get my head out of the sand.
Last report by naddy@
o clang:
- add back kernel printf %b length specifier support (%llb, etc, lost in
the update to 8.0.0)
o lld:
- Restore previous section after setting the MIPS ABI marker
- Fix output section alignement when entry size isn't a power of two
o arm64, amd64:
- Do not store the retguard cookie in frame in leaf functions if possible
- Emit variable length trap padding in retguard epilogue
o amd64:
- move code that selects retpoline by default to a different source file
o mips64:
- Fix a bug in memory operand handling
- Implement SGE pseudo-instructions
- Implement .cplocal directive
- Fix instruction guard
- Implement the 'h' register constraint on mips64
o sparc64:
- Remove cast that truncates immediate operands to 32 bits
I botched the library search paths when merging OpenBSD.cpp, repair
that. Also always add -L${LOCALBASE}/lib like was done before.
Looks like I committed an old diff which lacked SHARED_LIBS+=OptRemarks.
While here, force-disable Z3 support from math/z3.
Most patch merges inspired from patrick's work on base-clang. Consumer
ports fixed/marked BROKEN as discussed with bcallah@ and rsadowski@.
powerpc tests and PLIST fixes from cwen@, thanks! rust-1.36 tests by
semarie@
where we need a way to get to gcc-libs/libestdc++. This happened to
work so far except when it didn't (slight differences between built
and already installed llvm packages, maybe?), as found the hard way by
cwen@.
Our COMPILER support automatically amends WANTLIB and LIB_DEPENDS, so
use those to explicitely set:
- WANTLIB-main and WANTLIB-lldb
- LIB_DEPENDS-main and LIB_DEPENDS-lldb
The clang-tools-extra package contains helpful developer tools using
Clang's APIs.
Initial diff from me, further work by jca@ with feedback from sthen@.
OK jca@ and @sthen
* switch gcc4 module to gcc/8
* register the gcc/8 subpackages as updates for their gcc/4.9 counterparts
* sync the gcc version in devel/llvm
* add a quirk entry that gcj is obsolete
* unhook 4.9 from the build
Preparatory work by many.
ok sthen@
All the work done by Brad except for some fixes from base by mortimer@
and patrick@
Tests: amd64 Brad, jsg@ and me, i386: Brad, mips64: visa@, sparc64 me.
No tests on powerpc but a test build with only the PowerPC backend
packages on amd64.
- don't build clang itself with -fno-ret-protector and -mno-retpoline,
to regain performance
- use -mretpoline by default on amd64
- fix printf %b length specifiers
from Brad
