[8.5.0-3.0.2] - Fix Orabug 33451471 and backport CTF/BTF enhancements ctfc: Free CTF container elements in ctfc_delete_container ctf: Do not warn for CTF not supported for GNU GIMPLE ICE in btf_finalize when compiling with -gbtf Reviewed-by: Jose E. Marchesi [8.5.0-3.0.1] - Merge the following oracle patches to OL8.5 beta: - Fix an aarch64 compilation error triggered by the oracle patch gcc9-multiple-changes-align.patch on OL8U5 source base. gcc-fix-aarch64-tune-params.patch - Fix Orabug 33281392 Update CTF and BTF support in OL8 GCC This commit brings the support for CTF/BTF debug formats at par with upstream. GCC now generates the CTF/BTF debug information by using the internal DWARF representation. For backward compatibility reasons, OL8 GCC continues to support -gt command line option. - Add complex divide improvement backport of upstream commit 54f0224d55a1b56dde092460ddf76913670e6efc - Fix Orabug 32301371 - bug using gcov with preserve paths option This is the same bug as GCC bug PR gcov-profile/88994 gcc9-pr88994.patch - Fix generation of CTF type records for completed structs referred thru pointers. Orabug 31095790. - Fix Orabug 29838827 - provide an option to adjust the maximum depth of nested #include This is the same bug as gcc upstream PR90581 from Gcc9: gcc9-pr90581.patch - Fix Orabug 29541051 - confusing error message when there is a problem with ASAN_OPTIONS "ERROR: expected "="" This is the same bug as gcc upstream PR89832 from Gcc9: gcc9-pr89832.patch - Update support for CTF Fix Orabug 30833294 GCC generates incorrect CTF for single element arrays Fix Orabug 30808764 CTF generation fails when __attribute__ is used - Apply ares/neoverse support patches only ifarch aarch64. - Add 4 patches from gcc9 to support Arm Ares and Neoverse-N1 for Aarch64 gcc9-add-vec-reverse.patch gcc9-multiple-changes-align.patch gcc9-initial-mcpu-ares-tuning.patch gcc9-add-support-for-neoverse-n1.patch - Update support for CTF Fix Orabug 30778534 gcc should generate CTF for functions at file-scope only Fix Orabug 30779193 CTF generation fails for some flavors of vla Fix Orabug 30784275 Fix issues wtih CTF generation for typedef constructs ctf-3-generation-and-emission-for-a-single-compilation.patch ctf-4-update-ctf-testsuite.patch - Add support for CTF in GCC Fix Orabug 30102948 gcc: Add CTF generation to compiler Fix Orabug 30102949 gcc: Add CTF generation to compiler ctf-1-new-function-lang_GNU_GIMPLE.patch ctf-2-command-line-options-gtLEVEL.patch ctf-3-generation-and-emission-for-a-single-compilation.patch ctf-4-update-ctf-testsuite.patch ctf-5-handle-ctf-sections-when-lto-enabled.patch - CVE-2018-12207 / Intel SKX102 OL8 gcc: Intel Mitigation for CVE: CVE-2018-12207 - Allow -flto -Wa,-mbranches-within-32B-boundaries to pass -mbranches-within-32B-boundaries to GNU assembler. Without -lfto, -Wa,-mbranches-within-32B-boundaries to pass -mbranches-within-32B-boundaries to GNU assembler using existing GCC binaries. - Mitigation patch: gcc8-Fix-Wa-with-flto.patch - Fix Orabug 29968294 - Heap corruption with fprofile-dir=%p prevents profiling parallel processes, needed for RDBMS: Add patch to fix PR86057 from Gcc9: gcc9-pr86057.patch - Fix Orabug 30044244 - Profile directory concatenated with object file path This is the same bug as gcc upstream PR91971: gcc9-pr85759.patch gcc10-pr91971.patch - Fix Orabug 29599147 - Need -fprofile-dir=%q{VAR} backported to gcc8 This is the similar GCC PR47618, add the fix from GCC9: gcc9-pr47618.patch - Fix Orabug 29272977 - DB SUPPORT: Need way to dump inlining report from GCC Add -fopt-info-inline support from GCC9: gcc9-opt-info-inline.patch - Fix Orabug 29273006 - DB SUPPORT: need way to turn off inlining of global functions Add -flive-patching support from GCC9: gcc9-fipa-reference-addressable.patch gcc9-fipa-stack-alignment.patch gcc9-add-fomit-frame-pointer-to-test.patch gcc9-extend-live-patching-option-handling.patch gcc9-ipa-stack-alignment-386-test.patch - Introduce "oracle_release" into .spec file. Echo it to gcc/DEV-PHASE. - Backport 17 ampere patches from https://git.theobroma-systems.com/ampere-computing/gcc.git/log/?h=gcc-8_2_0-amp3-branch e18301133ea622f6d6796ded1d15466e70475cf8: Retpoline for aarch64. d735f3ae4712f66362326d179b4d7e9332c79677: Revert 2017-10-24 Richard Biener 271e2811e59c0c77fc022fa86a7030f20b4cac8e: Correct the maximum shift amount for shifted 0512749950d927de3dd695f2f2aacdfd30cf32fd: Add CPU support for Ampere Computings eMAG. c8b87078f9e0714cb9cab602e12a18ceb12df05a: eMAG/Xgene: Procedural cost-model for X-Gene 74610471b3577c5d465c3fd095a65b796b1e074c: Updating cost table for xgene1. ddba1553ac412be5596e6e2962c148032c4cf231: [AArch64] Add Xgene1 prefetch tunings. b7ebb0a10a8900324074070188a0936ed81b28a4: [AArch64] Fix in xgene1_addrcost_table 393dc5c50d55d069f91627bf0be5bab812978850: X-Gene: Adapt tuning struct for GCC 8. b9136d58824af2118c4969c3edb42cad3318b08f: tree-ssa-list-find-pipeline: Add pipelining loads for list finds. 095496dd8a9491a17a9caec173281ad02e559df5: uncse: Added pass to undo common subexpression elimination. a7c8dc238e3656e9d2f9256ee76f933c8d7956fb: loop-prefetcher: Adapt defaults for X-Gene cores. 256307f293f1750851576e14c8a42b696eced2da: tree-ssa-cpp: Dont crash on SSA names without definition stmts. 6e32f53be4f6733f6bfe267ad2337aecaf4047f6: Introduce new option -funroll-more. 1ac2485a2fced091a5cce6343fe6a6337f850e73: New option to bypass aliasing-checks. 66d7d833bece61e58998ad53a609cd32e3ee4fad: cfgloopmanip: Allow forced creation of loop preheaders. c4f89d50e200538b1ac8889801705300e0b27ef2: Add new pass to optimise loops. [8.5.0-3] - fix mangling of lambdas in default args - add a few Provides: bundled [8.5.0-2] - revert upstream PR85873 gcc-8 fix, apply the fix from gcc-9 - fix "this" adjustment for devirtualized call [8.5.0-1] - update from GCC 8.5 release - this includes a fix for PR target/87839