--- - branch: MAIN date: Fri Aug 12 08:37:58 UTC 2022 files: - new: '1.74' old: '1.73' path: pkgsrc/lang/llvm/Makefile pathrev: pkgsrc/lang/llvm/Makefile@1.74 type: modified - new: '1.20' old: '1.19' path: pkgsrc/lang/llvm/PLIST pathrev: pkgsrc/lang/llvm/PLIST@1.20 type: modified - new: '1.19' old: '1.18' path: pkgsrc/lang/llvm/buildlink3.mk pathrev: pkgsrc/lang/llvm/buildlink3.mk@1.19 type: modified - new: '1.35' old: '1.34' path: pkgsrc/lang/llvm/distinfo pathrev: pkgsrc/lang/llvm/distinfo@1.35 type: modified - new: '1.14' old: '1.13' path: pkgsrc/lang/llvm/options.mk pathrev: pkgsrc/lang/llvm/options.mk@1.14 type: modified - new: '1.9' old: '1.8' path: pkgsrc/lang/llvm/version.mk pathrev: pkgsrc/lang/llvm/version.mk@1.9 type: modified - new: '1.7' old: '1.6' path: pkgsrc/lang/llvm/patches/patch-cmake_config-ix.cmake pathrev: pkgsrc/lang/llvm/patches/patch-cmake_config-ix.cmake@1.7 type: modified id: 20220812T083758Z.bbf5dc71f21e5ec450c3720fd0ee4f7b1cd2b280 log: "lang/llvm: update to 14.0.6\n\nUpdated in wip by @wiz and myself.\n\n14.0.6\n-This release is a bug-fix release and is API and ABI compatible with 14.0.0.\n This release fixes a bug that was introduced in 14.0.5.\n This will likely be the last 14.0.x release.\n\n14.0.5\n-This release is a bug-fix release and is API and ABI compatible with 14.0.0.\n Note that we have adopted a new release schedule for the project, so there\n will be a new bug-fix release (14.0.x) every 2 weeks.\n 14.0.5 is the last planned release, but we may do a 14.0.6 release if there\n are critical issues found in 14.0.5.\n\n14.0.4\n-This release is a bug-fix release and is API and ABI compatible with 14.0.0.\n\n14.0.0\n\nChanges to the LLVM IR\n \ Using the legacy pass manager for the optimization pipeline is deprecated\n \ and will be removed after LLVM 14. In the meantime, only minimal effort will\n \ be made to maintain the legacy pass manager for the optimization pipeline.\n\n \ Max allowed integer type was reduced from 2^24-1 bits to 2^23 bits.\n\n Max allowed alignment was increased from 2^29 to 2^32.\n\nChanges to building LLVM\n \ Building LLVM with Visual Studio now requires version 2019 or later.\n\nChanges to the AArch64 Backend\n Added support for the Armv9-A, Armv9.1-A and Armv9.2-A architectures.\n\n The compiler now recognises the 窶徼une-cpu窶� function attribute to support\n the use of the -mtune frontend flag. This allows certain scheduling features\n and optimisations to be enabled independently of the architecture.\n If the 窶徼une-cpu窶� attribute is absent it tunes according to the\n 窶徼arget-cpu窶�.\n\n Fixed relocations against temporary symbols (e.g. in jump tables and\n constant pools) in large COFF object files.\n\n \ Auto-vectorization now targets SVE by default when available.\n\nChanges to the ARM Backend\n Added support for the Armv9-A, Armv9.1-A and Armv9.2-A architectures.\n\n \ Added support for the Armv8.1-M PACBTI-M extension.\n\n Changed the assembly comment string for MSVC targets to @ (consistent with\n the MinGW and ELF targets), freeing up ; to be used as statement separator.\n\nChanges to the PowerPC Target\nLinux improvements:\n Provided a number of builtins for compatibility with the XL compiler.\n\n Allow MMA builtin types in pre-P10 compilation units.\n\n Add support for Return Oriented Programming (ROP) protection for 32 bit.\n\n Refactored code to use more inclusive language.\n\n Switched to LLD as the default linker for pre-built Linux binaries.\n\n Enabled IEEE quad long double on Linux via PPC_LINUX_DEFAULT_IEEELONGDOUBLE\n in cmake config.\n\n Added __ibm128 type to represent IBM double-double format, also\n available as __attribute__((mode(IF))).\n\n \ -mfloat128 can now be used in Linux subtargets with VSX enabled.\n\n Added quadword atomic load/store support in codegen; not enabled by default.\n\n Codegen improvements for splat load, byval parameter, stack lowering, etc.\n\n Implemented P10 instruction scheduling model.\n\n Implemented P10 instruction fusion pairs.\n\n \ Improved handling of #pragma clang loop unroll_and_jam.\n\n Various bug fixes.\n\nAIX Support/improvements:\n Variadic (ellipsis) functions with C complex types are now supported.\n\n Added toc-data support for AIX 64-bit.\n\n \ Added toc-data support for read-only globals.\n\n Updated default target on AIX from pwr4 to pwr7.\n\n AIX 64-bit code generation now uses fast-isel for O0.\n\n Added DWARF support for 32-bit XCOFF.\n\nChanges to the RISC-V Target\n Codegen improvements for RV64 around the selection of addw/subw/mulw/slliw\n \ instructions and removal of redundant sext.w instructions (using the new\n \ RISCVSExtWRemoval pass).\n\n The various RISC-V vector extensions were updated to version 1.0 and are no\n longer experimental.\n\n The Zba, Zbb, Zbc, and Zbs bit-manipulation extensions were updated to\n version 1.0 and are no longer experimental.\n\n Added MC layer support for the ratified scalar cryptography extensions.\n\n The Zfh and Zfhmin extensions for half-precision floating point were updated\n to version 1.0 and are no longer experimental.\n\n \ Added support for the .insn directive.\n\n Various improvements to immediate materialisation, including when\n bit-manipulation extensions are enabled. Additionally, the constant pool is\n now used for large integers.\n\n Added support for constrained FP intrinsics for scalar types.\n\n Added support for CSRs introduced in the Sscofpmf, Smstateen, and Sstc\n extensions.\n\n The experimental 窶å\N­bproposedc窶� extension was removed, as was the 窶é\x8C¬çª¶ï¿½\n \ extension (including all bit-manipulation sub-extensions). Individual 窶å\N­b*窶�\n \ extensions should be used instead.\n\nChanges to the X86 Target\n Support for AVX512-FP16 instructions has been added.\n\n Removed incomplete support for Intel MPX. (D111517)\n\nChanges to the AMDGPU Target\n\nChanges to the Windows Target\n\n Changed how the .pdata sections refer to the code they窶å\x86³e describing, to\n avoid conflicting unwind info if weak symbols are overridden.\n\n \ Fixed code generation for calling support routines for converting 128 bit\n \ integers from/to floats on x86_64.\n\n The preferred path separator form (backslashes or forward slashes) can be\n configured in Windows builds of LLVM now, with the\n LLVM_WINDOWS_PREFER_FORWARD_SLASH CMake option. This defaults to true in\n MinGW builds of LLVM.\n\n Set proper COFF symbol types for function aliases (e.g. for Itanium C++\n constructors), making sure that GNU ld exports all of them correctly as\n functions, not data, when linking a DLL.\n\n \ Handling of temporary files on more uncommon file systems (network mounts,\n \ ramdisks) on Windows is fixed now (which previously either errored out or\n \ left stray files behind).\n\nChanges to the C API\n\n LLVMSetInstDebugLocation has been deprecated in favor of the more general\n LLVMAddMetadataToInst.\n\n \ Fixed building LLVM-C.dll for i386 targets with MSVC, which had been broken\n \ since the LLVM 8.0.0 release.\n\nChanges to the LLVM tools\n\n llvm-cov: -name-allowlist is now accepted in addition to -name-whitelist.\n -name-whitelist is marked as deprecated and to be removed in future\n releases.\n\n llvm-ar now supports --thin for creating a thin archive. The modifier T has\n a different meaning in some ar implementations. (D116979)\n\n llvm-ar now supports reading big archives for XCOFF. (D111889)\n\n llvm-nm now demangles Rust symbols. (D111937)\n\n \ llvm-objcopy窶å\x86± ELF port now avoids reordering section headers to preserve\n \ st_shndx fields of dynamic symbols. (D107653)\n\n llvm-objcopy now supports --update-section for ELF and Mach-O. (D112116)\n (D117281)\n\n llvm-objcopy now supports --subsystem for PE/COFF. (D116556)\n\n llvm-objcopy now supports mips64le relocations for ELF. (D115635)\n\n llvm-objcopy --rename-section now renames relocation sections together with\n their targets. (D110352)\n\n llvm-objdump --symbolize-operands now supports PowerPC. (D114492)\n\n llvm-objdump -p now dumps PE header. (D113356)\n\n llvm-objdump -R now supports ELF position-dependent executables. (D110595)\n\n llvm-objdump -T now prints symbol versions. (D108097)\n\n \ llvm-readobj: Improved printing of symbols in Windows unwind data.\n\n llvm-readobj now supports --elf-output-style=JSON for JSON output and\n --pretty-print for pretty printing of this output. (D114225)\n\n llvm-readobj now supports several dump styles (--needed-libs, --relocs,\n --syms) for XCOFF.\n\n llvm-symbolizer now supports 窶é\x9B»ebuginfod. (D113717)\n\n llvm-cov now accepts 窶å½\x97llowlist窶� spelling for -name-allowlist.\n\n llvm-nm now supports XCOFF object files.\n\n \ Added --needed-libs, aux header, and symbols support in llvm-readobj.\n\n \ Added --symbolize-operands support in llvm-objdump.\n\n Tools that read archive files now support reading AIX big format archive\n files.\n\n Added dump section support in obj2yaml.\n\n Added yaml2obj support for 64-bit XCOFF.\n\nChanges to LLDB\n\n A change in Clang窶å\x86± type printing has changed the way LLDB names array types\n (from int [N] to int[N]) - LLDB pretty printer type name matching code may\n need to be updated to handle this.\n\n The following commands now ignore non-address bits (e.g. AArch64 pointer\n signatures) in address arguments. In addition, non-address bits will not be\n shown in the output of the commands.\n\n memory find\n\n memory read\n\n memory region (see below)\n\n memory tag read\n\n memory tag write\n\n \ The memory region command and GetMemoryRegionInfo API method now ignore\n \ non-address bits in the address parameter. This also means that on systems\n \ with non-address bits the last (usually unmapped) memory region will not\n \ extend to 0xF窶ｦF. Instead it will end at the end of the mappable range that\n the virtual address size allows.\n\n The memory read command has a new option --show-tags. Use this option to\n show memory tags beside the contents of tagged memory ranges.\n\n Fixed continuing from breakpoints and singlestepping on Windows on ARM/ARM64.\n\n LLDB has been included in Windows on ARM64 binary release with Python\n support disabled.\n\nChanges to BOLT\n\n \ BOLT project is added to the LLVM monorepo. BOLT is a post-link optimizer\n \ developed to speed up large applications. Build and usage instructions are\n \ given in README.\n\nAdditional Information\n\nA wide variety of additional information is available on the LLVM web page, in\nparticular in the documentation section. The web page also contains versions of\nthe API documentation which is up-to-date with the Git version of the source\ncode. You can access versions of these documents specific to this release by\ngoing into the llvm/docs/ directory in the LLVM tree.\n\nIf you have any questions or comments about LLVM, please feel free to contact\nus via the mailing lists.\n" module: pkgsrc subject: 'CVS commit: pkgsrc/lang/llvm' unixtime: '1660293478' user: pin