)]}' { "log": [ { "commit": "6333d1ff93a97c8071e7393abc71da23b5a01706", "tree": "f7af3e00ac6c42e01d342a1e11d8c740b0f06e53", "parents": [ "110d383443c32ef09cf01f6cbf51cf91bc43101f" ], "author": { "name": "Andrew Haberlandt", "email": "ahaberlandt@apple.com", "time": "Sat Jul 11 01:40:02 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Jul 11 01:45:28 2026" }, "message": "[Darwin] Mark scalbn tests unsupported on macOS 26.4 (#208858)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 43ef833f39af617eea9850a1818d74ac376296ae\n" }, { "commit": "110d383443c32ef09cf01f6cbf51cf91bc43101f", "tree": "54f3d9324df45f974da9b695acec2ae9826901cb", "parents": [ "f50b2fb0b4fe6d058401f5c383b043506d30f893" ], "author": { "name": "Mike Kruskal", "email": "mkruskal@google.com", "time": "Fri Jul 10 22:43:12 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 10 22:45:25 2026" }, "message": "[compiler-rt] [sanitizer_common] Add a testcase reproducing SIGSEGV in ForEachMappedRegion (#208358)\n\nLikely related to https://github.com/llvm/llvm-project/issues/84482 and\nhttps://github.com/llvm/llvm-project/issues/21068\n\nAssisted-by: Gemini\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5a73020f0261922e1f30cd458f7d0bd757fb9ba1\n" }, { "commit": "f50b2fb0b4fe6d058401f5c383b043506d30f893", "tree": "17ff54e54617c15107d2472d80bafa462f7ab1ed", "parents": [ "48dd6a863d506c5a8c324bba9e28fcc08fbdd90d" ], "author": { "name": "Kyungtak Woo", "email": "kevinwkt@google.com", "time": "Fri Jul 10 20:28:34 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 10 20:30:56 2026" }, "message": "[compiler-rt][hexagon][test] Guard __hexagon_fmadf5 unit test with __hexagon__ define (#208818)\n\nWhen `compiler-rt/test/builtins/Unit/` C test files are compiled outside\nof a standard `lit` pipeline (such as direct standalone compilation or\ndownstream test harnesses where each `.c` file is built and run as an\nindependent executable against `libclang_rt.builtins.a`),\n`hexagon_fmadf5_test.c` unconditionally referenced `__hexagon_fmadf5`.\nBecause `__hexagon_fmadf5` is only provided in\n`libclang_rt.builtins-hexagon.a`, non-Hexagon builds failed at link\ntime.\n\nThis change is consistent with other architecture-specific builtin unit\ntests (such as `divdf3vfp_test.c`, `adddf3vfp_test.c`, and ARM/PPC\nbuiltin tests), this change wraps Hexagon-specific declarations,\nhelpers, and test logic inside `#if defined(__hexagon__)`. On\nnon-Hexagon targets, `main()` prints `\"skipped\\n\"` and returns `0`.\n\nIntroduced by https://github.com/llvm/llvm-project/pull/207373\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 37593bf3c58e693fb3bdaa2f5e7ab136a5ddbe75\n" }, { "commit": "48dd6a863d506c5a8c324bba9e28fcc08fbdd90d", "tree": "08874eeda162cd475eb48c036641f80bbc96fa3a", "parents": [ "4d7c41385c757a5890d6523e2efd32e234d25c91" ], "author": { "name": "Ian Anderson", "email": "iana@apple.com", "time": "Fri Jul 10 17:50:37 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 10 17:55:21 2026" }, "message": "[asan][test] Fix Posix/new_array_cookie_with_new_from_class.cpp to work for arm targets (#208378)\n\narm targets, and arm64 targets on Darwin, use a 2-pointer array cookie\nsize. Update the test to support that.\n\nAssisted-by: Claude Code\n\nrdar://181775880\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: cc0f75682622778349cef748d2a8d1dfd09e8185\n" }, { "commit": "4d7c41385c757a5890d6523e2efd32e234d25c91", "tree": "72e3453ee3975c3f68dce2f1d36559fb75f582ec", "parents": [ "ca6fb80108d0f077444d6ce216e04e40806f6bbc" ], "author": { "name": "Vladislav Dzhidzhoev", "email": "vdzhidzhoev@accesssoftek.com", "time": "Fri Jul 10 15:32:16 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 10 15:35:54 2026" }, "message": "[DebugMetadata][DwarfDebug][CodeView] Support function-local static variables in lexical block scopes (6/7) (#187927)\n\nRFC\nhttps://discourse.llvm.org/t/rfc-dwarfdebug-fix-and-improve-handling-imported-entities-types-and-static-local-in-subprogram-and-lexical-block-scopes/68544\n\nSimilar to imported declarations and types, the patch tracks\nfunction-local static variables (globals in terms of LLVM IR) in\nDISubprogram\u0027s \u0027retainedNodes\u0027. DwarfDebug is adjusted in accordance\nwith the aforementioned metadata change and provided a support of static\nlocals scoped by a lexical block. CodeViewDebug is modified to collect\nglobal variables from DISubprogram\u0027s \u0027retainedNodes\u0027 too.\n\nThe patch assumes that DICompileUnit\u0027s \u0027globals\u0027 no longer tracks static\nlocals and DwarfDebug would assert if any locally-scoped variables get\nplaced there.\n\nThis change has already been reviewed in\nhttps://reviews.llvm.org/D144008, but I’m opening a new PR because it\nhas been a while since it was approved. Additional logic was added to\nCloneFunction to deal with DISubprograms that have static locals: if\nsuch a subprogram is cloned, DIGlobalVariableExpressions are removed\nfrom retainedNodes of the clone. This is consistent with the current\nbehavior, in which DIGlobalVariableExpressions are not cloned.\n\nAuthored-by: Kristina Bessonova \u003ckbessonova@accesssoftek.com\u003e\nCo-authored-by: Tobias Gysi \u003ctobias.gysi@nextsilicon.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: d58b0562a340b043ed83b0fb4375e89ed74fc621\n" }, { "commit": "ca6fb80108d0f077444d6ce216e04e40806f6bbc", "tree": "c5b95cdd010a09080d62c198265e944a245ab27d", "parents": [ "499a9c8fe6de00d54fd0756fca5b9e8c8c375963" ], "author": { "name": "Zain", "email": "beingzainsv@gmail.com", "time": "Fri Jul 10 14:20:46 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 10 14:25:53 2026" }, "message": "[compiler-rt][hexagon] Fix `__hexagon_fmadf5` overflow under directed rounding (#207373)\n\nThe overflow path in `__hexagon_fmadf5` computed the round-mode\npredicate (inf vs. largest-finite) into p0, but then overwrote p0 with\n`dfcmp.eq(ATMP,ATMP)` before consuming it. Since ATMP is +inf,\n`dfcmp.eq(ATMP,ATMP)` is always true, so CTMP was unconditionally set to\ninfinity and the rounding-mode decision was discarded.\n\nAs a result, a finite-magnitude overflow returned +/-inf in every\nrounding mode.\n\nFix: Fuse the selection into a single packet so the round-mode predicate\nselects inf-vs-DBL_MAX, and recompute p0 \u003d dfcmp.eq afterwards only to\nraise the exception flags which matching the correct sequence already\nused\nby dfdiv.S.\n\nAlso, I\u0027ve added a regression test (`hexagon_fmadf5_test.c`) covering\nFMA special\ncases: NaN propagation, invalid operations (0*inf, inf+(-inf)), infinity\npropagation, exact results, signed-zero of exact sums, the single-\nrounding property, overflow under all rounding modes, and subnormal\nresults.\n\nPlease let me know if this test for the builtin is in the right place,\nor if it should be in some place else.\n\nAssisted by: GPT 5.5\n\nSigned-off-by: Zain Siddavatam \u003czsiddava@qti.qualcomm.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 8544134f118dcbd1543e29c1ae6b5ac019d4a138\n" }, { "commit": "499a9c8fe6de00d54fd0756fca5b9e8c8c375963", "tree": "919bbd668a75967357c5a58fbdd2bb124b2058eb", "parents": [ "1f6d207f8cdf712bb9e690b98d9e40ffb0ff361b" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Thu Jul 09 20:08:28 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jul 09 20:10:40 2026" }, "message": "[scudo] Small refactor of getTSDXXX functions. (#208127)\n\nSmall modification in getTSDAndLock() to have a single shared TSD avoid\nthe tryLock call. Also, remove some unnecessary comments.\n\nAdd a DCHECK in getTSDSlow and only check for \u003d\u003d 1 since a 0 number of\nshared TSDs is not supported.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 766161cd2fbe851bf1a67abe469269057aa854de\n" }, { "commit": "1f6d207f8cdf712bb9e690b98d9e40ffb0ff361b", "tree": "4d76289ffc4ecfb7d1f269c2ed43de0e9cadeb32", "parents": [ "65b1f193f33193b0321493965edbb7acb7170a09" ], "author": { "name": "David CARLIER", "email": "devnexen@gmail.com", "time": "Thu Jul 09 16:19:54 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jul 09 16:25:52 2026" }, "message": "[compiler-rt][memprof] fix off-by-one in shadow access-count accumula… (#208376)\n\n…tion.\n\nGetShadowCount and GetShadowCountHistogram used an inclusive loop up to\nMEM_TO_SHADOW(p + size), summing an extra adjacent cell when p + size is\ngranule-aligned. Use p + size - 1 to stay within the allocation.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: ef84e19e3fd3f13ffbc87544054e6605226dfc79\n" }, { "commit": "65b1f193f33193b0321493965edbb7acb7170a09", "tree": "e073d04c1f46b3faff530848c9e20cda4c82faa4", "parents": [ "ac301289387e544f372994efd82f178bf3b413e6" ], "author": { "name": "Mikołaj Piróg", "email": "mikolaj.maciej.pirog@intel.com", "time": "Tue Jul 07 13:03:15 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jul 07 13:05:28 2026" }, "message": "[X86] Sync compiler-rt cpu subtypes and vendors with libgcc and refactor internal cpu type tables in LLVM (#171172)\n\nThis is a continuation of previous PR:\nhttps://github.com/llvm/llvm-project/pull/168750\n\ncompiler-rt was synced with libgcc on ProcessorVendor and\nProcessorSubtype fields and so was llvm. Cpu type, subtype and vendor\nentries in X86TargetParser.def were refactored to use ABI_VALUE.\n\nLLVM doesn\u0027t set the ABI_VALUE for its enums -- clang now takes care of\nthat by reading the ABI_VALUE.\n\nI\u0027ve removed and added some comments to better explain what is going on.\n\nWhile at it, I\u0027ve added tests to check that right values are passed for\nvendor, subtypes, types and features (the last ones weren\u0027t added in\nfeature sync PR: https://github.com/llvm/llvm-project/pull/168750)\n\nParts of the PR (test) have been realized using Claude Code\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 91578461e8a112941afeed18679efb168b81f796\n" }, { "commit": "ac301289387e544f372994efd82f178bf3b413e6", "tree": "c303613cedc74abef05ea075aea8b208019a84cb", "parents": [ "230acc60848fc766159e2cd00e92a50856f19ed1" ], "author": { "name": "Jaydeep Chauhan", "email": "chauhan.jaydeep.ashwinbhai@intel.com", "time": "Tue Jul 07 07:43:27 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jul 07 07:45:53 2026" }, "message": "[X86] Remove AMX-TF32 (#207673)\n\nPer Intel Architecture Instruction Set Extensions Programming Reference\nrev. 62 https://cdrdv2.intel.com/v1/dl/getContent/922690 Revision\nHistory entry for revision -062\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5bca42d7d7cd55642e9e63fa3b4da11245d025c3\n" }, { "commit": "230acc60848fc766159e2cd00e92a50856f19ed1", "tree": "64752092ba8af9d6de03a6b0c95a87ba245acaee", "parents": [ "7d90b0b3ef61814d9c37d5d3695b8e3dd1a40375" ], "author": { "name": "Arthur Eubanks", "email": "aeubanks@google.com", "time": "Mon Jul 06 21:46:06 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jul 06 21:50:49 2026" }, "message": "Revert \"[Clang][Sema] Add -Wstringop-overread warning for source buffer overreads (#183004)\" (#207840)\n\nThis reverts commit 86f2e71cb8d165b59ad31a442b2391e23826133e.\n\nAnd revert followup ab1ef459069d1dbf4ee9e5a802537ababe294b25 (#205201).\n\nCauses crashes:\nhttps://github.com/llvm/llvm-project/pull/183004#issuecomment-4897195939\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 94b325469b20255d31879c80204800510ac4fd66\n" }, { "commit": "7d90b0b3ef61814d9c37d5d3695b8e3dd1a40375", "tree": "53366339ab192287eb4c6fb14a579022dbf18b4a", "parents": [ "80c8e5f688993c0c6c444a054c860a26c75bc54a" ], "author": { "name": "Aliaksei Urbanski", "email": "aliaksei.urbanski@gmail.com", "time": "Mon Jul 06 18:07:45 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jul 06 18:10:44 2026" }, "message": "[NFC] Fix \"is is\" typos (#206288)\n\nThese changes aim to fix common \"is is\" typos.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5e265484030eb034a14f4e4fec5abb0209605787\n" }, { "commit": "80c8e5f688993c0c6c444a054c860a26c75bc54a", "tree": "1870e13e08f1c927ed8fcf2c33b5fc9f537e0c89", "parents": [ "4290ea643c5b726177c079babf424245b66b9efa" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Mon Jul 06 17:54:14 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jul 06 17:56:13 2026" }, "message": "[scudo] Make death tests use SCUDO_XXX_DEATH_TEST. (#206196)\n\nOn Android, we want to disable the tombstone generation on expected\ncrashes to speed up running the tests. Also, this means that any\ntombstone files will be real crashes and it\u0027s not necessary to search\nfor the right tombstone if one of the tests crashes unexpectedly.\n\nIn order to do the above, add SCUDO_EXPECT_DEATH_TEST and\nSCUDO_ASSERT_DEATH_TEST that will disable tombstoned on Android but do\nnothing on all other platforms.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 96904e83cf69b0a02a5fb7dece9b258f562205e0\n" }, { "commit": "4290ea643c5b726177c079babf424245b66b9efa", "tree": "523fd008e779e2fd313bee43dba1f0037ca97c10", "parents": [ "7c0d0b69035bf68de58ce9605703511f21ce93b0" ], "author": { "name": "Aditya Medhane", "email": "sherlockedaditya@gmail.com", "time": "Mon Jul 06 13:32:40 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jul 06 13:35:44 2026" }, "message": "[scudo] Remove internal linkage from LargeBlock header helpers (NFC) (#207708)\n\n`addHeaderTag` and `getHeader` are static function templates in\n`secondary.h,` so any TU that includes the header without using them\ntrips `-Wunused-template`. Dropping static fixes the warning.\n\nPart of #202945.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: dcf969eaf2c25048542d13946ca5ff65be5bc7ab\n" }, { "commit": "7c0d0b69035bf68de58ce9605703511f21ce93b0", "tree": "dc049312267c052e316efb3fff85ef6bcae6b6e5", "parents": [ "824fd2b076dbee9069fa939cb0b7eed435e814f9" ], "author": { "name": "Aditya Medhane", "email": "sherlockedaditya@gmail.com", "time": "Mon Jul 06 13:29:39 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jul 06 13:30:22 2026" }, "message": "[compiler-rt][sanitizer_common] Fix -Wunused-template in test helpers… (#207706)\n\n`Ident` is a static function template in a test header, so any test TU\nthat never calls it trips `-Wunused-template`. Drop static.\n\n`check_locked` in `sanitizer_mutex_test.cpp` has no callers.\n\nPart of #202945.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 7cb7f074b71e62d33b22ced07d860947fac08480\n" }, { "commit": "824fd2b076dbee9069fa939cb0b7eed435e814f9", "tree": "59341ff5945631eece48c39933d9432949d1429e", "parents": [ "221020f27b523cb989d2dd037425123e42c1d754" ], "author": { "name": "Aditya Medhane", "email": "sherlockedaditya@gmail.com", "time": "Sat Jul 04 05:57:59 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Jul 04 06:01:40 2026" }, "message": "[compiler-rt][nsan] Remove unused checkValueConsistency template (NFC) (#207441)\n\ncheckValueConsistency is a static function template in nsan.cpp with no\ncallers, so it trips -Wunused-template. Remove the dead code.\n\nPart of #202945\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: db0f427a25a4caae97679a4a0caf05aed5dd47f5\n" }, { "commit": "221020f27b523cb989d2dd037425123e42c1d754", "tree": "374dbcee498646824c8c5f2a6dcde7a51765955d", "parents": [ "92cdd9216af39a0b4da7c5bed1fb2d31bd4eb567" ], "author": { "name": "Aditya Medhane", "email": "sherlockedaditya@gmail.com", "time": "Fri Jul 03 17:42:14 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 03 18:44:48 2026" }, "message": "[compiler-rt][hwasan] Remove internal linkage from check templates (NFC) (#207430)\n\n`SigTrap`, `CheckAddress`, and `CheckAddressSized` are static function\ntemplates in `hwasan_checks.h`, so any TU that includes the header\nwithout instantiating them trips -Wunused-template. Dropping static\nclears the warning.\nPart of #202945\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 14d98081a574c3c5f342e2d2e958a514369fa2f2\n" }, { "commit": "92cdd9216af39a0b4da7c5bed1fb2d31bd4eb567", "tree": "632ef9e73a270edcd546e158d590e56ec9438adb", "parents": [ "0eb3dcad630550cdf8a38d73dd9693b995c09fcb" ], "author": { "name": "Aditya Medhane", "email": "sherlockedaditya@gmail.com", "time": "Fri Jul 03 16:28:37 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 03 18:44:40 2026" }, "message": "[compiler-rt][sanitizer_common] Remove internal linkage from RegisterFlag (NFC) (#206308)\n\nRegisterFlag is a static function template in a header, so every TU that\nincludes it without calling it trips `-Wunused-template`. Dropping\nstatic gives it normal external linkage and clears the warning.\n\nNFC. Part of #202945.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 554e32395cd552dd36d3511d177f01735937d4f9\n" }, { "commit": "0eb3dcad630550cdf8a38d73dd9693b995c09fcb", "tree": "2faa2262cd1ecc55f94807bf9bf2169daef26e70", "parents": [ "4b36ff6225212d114cb1ca83848628c188eb4f1a" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri Jul 03 01:07:59 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jul 03 18:44:28 2026" }, "message": "[PGO][HIP][NFC] Fix hipModuleGetGlobal -Wunused-function warning (#207293)\n\nThe functions trigger the warning on Windows (without elf.h) and is\nfatal under -Werror.\nFix by adding [[maybe_unused]]. Alternatively it could be moved inside\nthe existing __has_include(\u003celf.h\u003e) block,; however that would trigger\n-Wunused-but-set-global on pHipModuleGetGlobal.\nCurrent fix is minimal and can be removed once hipModuleGetGlobal is\nsupported without elf.h.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 604db5fc463d2e252c2575ff519f479c88cfe093\n" }, { "commit": "4b36ff6225212d114cb1ca83848628c188eb4f1a", "tree": "f92eb33860b2de0c34a43a201404a558138aee7f", "parents": [ "3e37f58194d95c88bf272d573b8037eea0073e69" ], "author": { "name": "hulxv", "email": "hulxxv@gmail.com", "time": "Thu Jul 02 17:00:31 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jul 02 17:05:32 2026" }, "message": "[compiler-rt][builtins] introduce libc math routines (#197950)\n\nIntroduce using libc math routines in compiler-rt builtins. this pr adds\n`COMPILER_RT_USE_LIBC_MATH` flag and uses math routines in basic\narithmetic operations for Quad-Precision values\n\nRFC:\nhttps://discourse.llvm.org/t/rfc-explore-using-llvm-libc-math-routines-for-compiler-rt-floating-point-builtins/89670\n\nPart of #197824\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: cb4e7fff7d0e7fc2bf116302567f21405f63a748\n" }, { "commit": "3e37f58194d95c88bf272d573b8037eea0073e69", "tree": "c85aa23a7ee1ecaa129c1a1d6663985bda3dd7ec", "parents": [ "e6531d662e1951d93c55573b2f6b9011c1e650b6" ], "author": { "name": "Azat Khuzhin", "email": "a3at.mail@gmail.com", "time": "Thu Jul 02 16:56:57 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jul 02 17:00:17 2026" }, "message": "[sanitizer] Raise kMaxAllowedMallocSize to 1TB on AArch64 (#206649)\n\nlsan capped AArch64 allocations at 4GB and msan at 8GB. These limits\nwere tied to the smaller address space of the 32-bit-style allocator.\n\nThis became possible after 8e0c1aaf6ba3 (\"[msan] Use\nSizeClassAllocator64 for AArch64\"), which (together with the 48-bit VMA\nrequirement) switched AArch64 to SizeClassAllocator64, so the limit can\nnow match the other 64-bit targets (1ULL \u003c\u003c 40).\n\nFollow-up for #89728\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 64b1d17103e590da4bcc2c008c2f01c4253ec245\n" }, { "commit": "e6531d662e1951d93c55573b2f6b9011c1e650b6", "tree": "4304ed90600c36b5413d4e9ab798bc8e152379e1", "parents": [ "19596f2793abb1a7f908555a4336952845337939" ], "author": { "name": "Ganesh", "email": "Ganesh.Gopalasubramanian@amd.com", "time": "Thu Jul 02 15:43:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jul 02 15:45:57 2026" }, "message": "[X86] Add AVX512BMM support for AMD Zen 6 (znver6) (#182556)\n\nThis patch adds support for AVX512BMM (Bit Matrix Multiply) instruction\nset extension for AMD Zen 6 processors.\n\nAVX512BMM includes three instructions:\n - VBITREVB: Bit reverse within each byte\n - VBMACOR: Bit matrix multiply with OR accumulation\n - VBMACXOR: Bit matrix multiply with XOR accumulation\nThe following implementations for AVX512BMM are added:\n - Define __AVX512BMM__ macro for znver6\n - avx512bmmintrin.h, avx512bmmvlintrin.h header files\n - Implement _mm_bitrev_epi8, _mm256_bitrev_epi8, _mm512_bitrev_epi8\n - Implement _mm256/512_bmacor16x16x16 and bmacxor intrinsics\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 4df00678d86feafefcd953db317c80241b691d53\n" }, { "commit": "19596f2793abb1a7f908555a4336952845337939", "tree": "f037ed764107f8d6151a45fe9294e927bb52c3ec", "parents": [ "816da5256859ab49d76b3bea1026d5f88fa411f6" ], "author": { "name": "Giovanni Ortuño", "email": "gd.ortuno@gmail.com", "time": "Wed Jul 01 21:32:38 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jul 01 21:35:58 2026" }, "message": "[libfuzzer] Improve error message for file not found (#207081)\n\nWe\u0027re seeing many instances of this issue on ClusterFuzz. This PR adds\nmore information to help debug the cause of the problem.\n\nSee https://crbug.com/529865766 for more context.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: f74953b929fefb8edef80f6e99d596b72b540c66\n" }, { "commit": "816da5256859ab49d76b3bea1026d5f88fa411f6", "tree": "f6ad655dd3e0f67d5cd78b3906931876bd58bcd9", "parents": [ "1fff913ee2bad3857cec24a8a209d7186f80c5cc" ], "author": { "name": "Petr Hosek", "email": "phosek@google.com", "time": "Wed Jul 01 06:14:09 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jul 01 06:15:36 2026" }, "message": "[compiler-rt] Support triples that start with thumb*- (#206848)\n\nthumb*- can be used as an alias for arm*- (implying -mthumb).\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 4c27981352753962ffeba8ba9a6d2d49b70a240b\n" }, { "commit": "1fff913ee2bad3857cec24a8a209d7186f80c5cc", "tree": "fe6da660a2445f201214c4e85def182ee96c69c6", "parents": [ "fd7078e542f1a353697816ec610f52ea1ffd9b68" ], "author": { "name": "Roman Beliaev", "email": "r.beliaev@ispras.ru", "time": "Tue Jun 30 15:55:16 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 30 16:00:33 2026" }, "message": "Reland \"[MC/DC][Coverage] Enable profile correlation for MC/DC\" (#204675)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 668803fa73bdf2f8d095162a1f81c000d18458c9\n" }, { "commit": "fd7078e542f1a353697816ec610f52ea1ffd9b68", "tree": "b6049c01d72baaa6e09d8b2ebd901d18054fb2de", "parents": [ "14f71a7b2ddfe993610abb70b4c45981e68dd230" ], "author": { "name": "maflcko", "email": "6399679+maflcko@users.noreply.github.com", "time": "Tue Jun 30 14:37:56 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 30 14:41:25 2026" }, "message": "compiler-rt: intercept fortified read/pread wrappers (#206228)\n\nAdd TSAN/common interceptors for glibc fortified read/pread calls, so\nthat they follow the same blocking and signal-handling path as the plain\nlibc symbols.\n\nThe regression test from https://github.com/llvm/llvm-project/pull/77789\nfor `read`, when compiled with `-D_FORTIFY_SOURCE\u003d3` fails without the\nnew interceptors.\n\nI have a regression test for pread(64) as well, but I am not sure if\nthere is much value in adding it.\n\nI\u0027ve added a read smoke test in sanitizer common.\n\n---------\n\nCo-authored-by: MarcoFalke \u003c*~\u003d`\u0027#}+{/-|\u0026$^_@721217.xyz\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: f801f9d0103dd715fde0d3518d47b5fe605f093c\n" }, { "commit": "14f71a7b2ddfe993610abb70b4c45981e68dd230", "tree": "1c30e4c0bff10e4d3fe17978cc9a5641653a3656", "parents": [ "f365260d3d2b19f5f61d356d99b5e2bb5d158193" ], "author": { "name": "Dan Blackwell", "email": "dan_blackwell@apple.com", "time": "Tue Jun 30 13:20:59 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 30 13:26:06 2026" }, "message": "[Darwin][TSan] Fix false positive race report when using dispatch_apply (#204866)\n\nCurrently I have seen some false positive race reports caused by\ninstrumented block copy code that writes after the release-annotation.\nWhen the block runs on another thread and reads the copied data, it\nincorrectly reports this as a race.\n\nThis patch casts the block to a `void *` in order to avoid the block\ncopy - thus avoiding the reported race. This is safe because\n`dispatch_apply` runs synchronously, so the block lives beyond all\ninvocations.\n\nrdar://92286127\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 00617c92e10fb55be86e6fcf60ef2f5717d8eb92\n" }, { "commit": "f365260d3d2b19f5f61d356d99b5e2bb5d158193", "tree": "e00db1dc6419bee457d0caced7d7d486ed0d7b1a", "parents": [ "0a4c4029c9b22a8a27b4b35558fd5dbe1479f3f4" ], "author": { "name": "Kelvin Li", "email": "kli@ca.ibm.com", "time": "Tue Jun 30 12:15:28 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 30 12:22:50 2026" }, "message": "[compiler-rt][test][builtins] Check linker support for --start-group (#205610)\n\nThis patch is to add a check for the linker whether the --start-group\noption is supported before passing to the linker. This can accommodate\ndifferent linker requirements.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: c2da3ace1beac5a7d94e58fd018bbce7298b660d\n" }, { "commit": "0a4c4029c9b22a8a27b4b35558fd5dbe1479f3f4", "tree": "b88f0a931abe84873a3d72b1afe2efa25743a91d", "parents": [ "963dd0538b67919ccd2f99a011a1402e109ea73b" ], "author": { "name": "Xiaomeng Zhang", "email": "zhangxiaomeng@hygon.cn", "time": "Tue Jun 30 06:45:55 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 30 06:50:40 2026" }, "message": "[X86] Hygon C86-4G-M8 Initial enablement (#204587)\n\nThis patch adds initial support for Hygon C86-4G-M8 architectures:\n\n- Added C86-4G-M8 CPU targets recognition in Clang and LLVM\n- Added C86-4G-M8 to target parser and host CPU detection\n- Updated compiler-rt CPU model detection for C86-4G-M8\n- Added C86-4G-M8 to various optimizer tests\n- Added scheduler models and llvm-mca tests for C86-4G-M8 CPU targets\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 77de767af99b151046d1f25df4e37ba61e79531c\n" }, { "commit": "963dd0538b67919ccd2f99a011a1402e109ea73b", "tree": "a796a3fcdf06d67a9d1965b67f8260af2972a7f0", "parents": [ "8d8a646c841f6d0a6c3186fed1a2fc1c46069588" ], "author": { "name": "Petr Hosek", "email": "phosek@google.com", "time": "Mon Jun 29 18:16:27 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 29 18:20:57 2026" }, "message": "[CMake] Warn when the version of CMake is older than 3.31.0 (#204203)\n\nThis is a preparation to raise the minimum required CMake version to\n3.31.0 in LLVM 24.\n\nSee\nhttps://discourse.llvm.org/t/rfc-raising-minimum-required-cmake-version-to-3-31/91086.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: e192ec42b426f6ffbc25419238a90f3c59088587\n" }, { "commit": "8d8a646c841f6d0a6c3186fed1a2fc1c46069588", "tree": "d589993d1447ceb07096679612dc700910426823", "parents": [ "6d4e3cb1d8b9a6a1d03d5352b7e7998cfdc63fde" ], "author": { "name": "Andrew Haberlandt", "email": "ahaberlandt@apple.com", "time": "Mon Jun 29 17:54:41 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 29 17:55:47 2026" }, "message": "[sanitizer_common] [Darwin] Update version mapping for macOS 27-aligned releases (#204608)\n\nThis updates `MapToMacos` et. al. to produce correct results for macOS\n26 and macOS 27-aligned releases.\n\nrdar://177997021\nrdar://169356606\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: d0c64764ebd4be00a02d449250fe0436fcf8d70a\n" }, { "commit": "6d4e3cb1d8b9a6a1d03d5352b7e7998cfdc63fde", "tree": "93b9c90a989cb1d40ff7d28742b0a55850c4545b", "parents": [ "57d6af57c374757d67349eaf85b41efc3c603bd3" ], "author": { "name": "Yaxun (Sam) Liu", "email": "yaxun.liu@amd.com", "time": "Mon Jun 29 15:30:45 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 29 15:35:30 2026" }, "message": "[PGO][AMDGPU] Add uniformity-aware offload profile format and instrumentation (#190708)\n\nAdd the producer-side format and instrumentation pieces that layer on\ntop\nof the basic HIP offload PGO support.\n\nThis extends the raw/indexed profile format with wave size and uniform\ncounters, teaches llvm-profdata to read and merge them, adds the AMDGPU\nsampling runtime, and emits block.uniformity.profile metadata during\nPGO instrumentation.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: ab95a8e7bb519a91abdfa5722ec40c0e0909f8f4\n" }, { "commit": "57d6af57c374757d67349eaf85b41efc3c603bd3", "tree": "4bf091b0981d77e6c6d20c0b04cb40451615ea39", "parents": [ "732884b9a81a50b2e9479813d109ba4780d990c0" ], "author": { "name": "Jakob Koschel", "email": "jakobkoschel@google.com", "time": "Mon Jun 29 11:39:17 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 29 11:40:12 2026" }, "message": "[SafeStack] Allocate unsafe sigaltstack (#206463)\n\nPR https://github.com/llvm/llvm-project/pull/196969 was approved and\nmerged but with `spr` the wrong base branch was set.\n\nThis merges the approved changes into `main`\n\n---------\n\nCo-authored-by: Paul Walker \u003cpaul.walker@arm.com\u003e\nCo-authored-by: Arseniy Obolenskiy \u003carseniy.obolenskiy@amd.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: acd0f6fa1eb5b5cf78ddbd3281cdcf2e1e9587d3\n" }, { "commit": "732884b9a81a50b2e9479813d109ba4780d990c0", "tree": "d4cd3ee9483738cb377b0accd9ec6c607dad23ec", "parents": [ "e97f9aa3a7f6d3b2f17b54dba83e930214fe1ce6" ], "author": { "name": "Simon Tatham", "email": "simon.tatham@arm.com", "time": "Fri Jun 26 06:49:05 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 26 06:50:25 2026" }, "message": "[compiler-rt][ARM] Optimized single precision FP add/sub (#179929)\n\nThis adds new implementations of single-precision add/sub in both Thumb1\nand Arm/Thumb2 assembler.\n\nBoth of the new implementations are included in the builtins library if\nthe `COMPILER_RT_ARM_OPTIMIZED_FP` cmake option is enabled (as it is by\ndefault).\n\nThere was already a Thumb1 assembler implementation of single-precision\nadd/sub, slower but also smaller. I\u0027ve kept it (although it\u0027s been moved\ninto the `thumb1` subdirectory), and if you _don\u0027t_ enable\n`COMPILER_RT_ARM_OPTIMIZED_FP`, it will be selected in place of the new\nversion.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: affc89f7cb6439bd3974ef816e8f9872e76166c1\n" }, { "commit": "e97f9aa3a7f6d3b2f17b54dba83e930214fe1ce6", "tree": "55be6e9a9b333d3c7d3425fe8772819b9c3a0f7b", "parents": [ "c1350cc43e13853c6f34a9d4fe736cd0392f2b01" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Fri Jun 26 02:38:14 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 26 02:41:02 2026" }, "message": "[Hexagon] Add ShadowCallStack support (#200508)\n\nImplement the software ShadowCallStack for Hexagon.\n\nOn Hexagon, r19 is used as the shadow stack pointer (reserved via\n-ffixed-r19). On function entry the LR (r31) is saved to the shadow\nstack and the pointer is advanced; on exit the LR is restored from the\nshadow stack before returning.\n\nPrologue sequence:\n\n r19 \u003d add(r19, #4)\n memw(r19+#-4) \u003d r31\n\nEpilogue sequence (between deallocframe/jumpr r31):\n\n r31 \u003d memw(r19+#-4)\n r19 \u003d add(r19, #-4)\n\nSCS is only emitted when hasFP() is true: the function uses allocframe.\nLeaf functions that never touch the stack do not need protection.\n\nWhen SCS is active the combined L4_return (deallocframe+jump) is not\nused; instead L2_deallocframe and PS_jmpret are kept separate so the\nshadow-stack restore can be inserted between them.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 2919ad75affbfae061415f451a4ba8b7c76d44a3\n" }, { "commit": "c1350cc43e13853c6f34a9d4fe736cd0392f2b01", "tree": "39623edb4ba59d5fa5e3559d25b7223f4e7c4f80", "parents": [ "3298b5ebaa53be38f2bd4981ef59b3771ebdd8a1" ], "author": { "name": "Johannes Doerfert", "email": "jdoerfert.llvm@gmail.com", "time": "Fri Jun 26 02:13:59 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 26 02:15:32 2026" }, "message": "Revert \"[Instrumentor] Add runtime examples: [1/N] A flop counter\" (#205960)\n\nThis reverts commit 61cbfabb7ade682a64f516c871b2bacb1e3e324a.\n\nFails compiler-rt standalone builds, though, locally it works fine :(\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3ddeaebfd33fe0a3e45883c21ab1eee44ced83a6\n" }, { "commit": "3298b5ebaa53be38f2bd4981ef59b3771ebdd8a1", "tree": "35cc64eec7c6fbc79b412b5ef0eddad45d2f174d", "parents": [ "0ebf65206ab44fde0ce81b4becf9c2e402a66b50" ], "author": { "name": "Johannes Doerfert", "email": "jdoerfert.llvm@gmail.com", "time": "Fri Jun 26 01:05:02 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 26 01:11:14 2026" }, "message": "[Instrumentor] Add runtime examples: [1/N] A flop counter (#205698)\n\nThis adds a instrumentor-tools folder into compiler RT to showcase use\ncases of the instrumentor. The initial example is a program that, via\ninstrumentation, counts the number of flops performed. Call and\nintrinsic support will follow after #198042.\n\nThis is the second try with more CMake magic after\nhttps://github.com/llvm/llvm-project/pull/205221 failed on some\nplatforms.\n\nPartially developped by Claude (AI), tested and verified by me.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: cc91bbc5d88e56b85cd80e391d7faddd422bb3e4\n" }, { "commit": "0ebf65206ab44fde0ce81b4becf9c2e402a66b50", "tree": "39623edb4ba59d5fa5e3559d25b7223f4e7c4f80", "parents": [ "ac5c7765d054db384dc0cfcd27f50b20398259c7" ], "author": { "name": "Matthew Nagy", "email": "matthew.nagy@sony.com", "time": "Thu Jun 25 12:33:58 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 25 12:35:58 2026" }, "message": "[TySan] Remove incorrect memcpy shadow memory tracking (#199958)\n\nFixes the false positive in\nhttps://github.com/llvm/llvm-project/issues/122934\n\nmemcpy is allowed to bypass strict aliasing rules (see\nhttps://en.cppreference.com/c/string/byte/memcpy) so we shouldn\u0027t alter\nshadow memory when it is used\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 1f84b4aab5c714ea8be7ee8f69d00c69b7cecaba\n" }, { "commit": "ac5c7765d054db384dc0cfcd27f50b20398259c7", "tree": "971208d9910f0df50c544b22ed6aff8ca98474a1", "parents": [ "5f6e36adeef9b4cc534fd0a6828916232d5e4cb8" ], "author": { "name": "Johannes Doerfert", "email": "jdoerfert.llvm@gmail.com", "time": "Wed Jun 24 23:43:40 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 23:45:45 2026" }, "message": "Revert \"[Instrumentor] Add runtime examples: [1/N] A flop counter (#205221)\" (#205696)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: b26636154374424e3d2ded25e853dbcc975640de\n" }, { "commit": "5f6e36adeef9b4cc534fd0a6828916232d5e4cb8", "tree": "f42d034aeeb3a77407ec6d9259e2cf4810c6f7e3", "parents": [ "17768d32b73335768e2a93ffb55841b8f4c64bde" ], "author": { "name": "Johannes Doerfert", "email": "jdoerfert.llvm@gmail.com", "time": "Wed Jun 24 23:25:10 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 23:30:23 2026" }, "message": "[Instrumentor][FIX] Ensure CXX headers are available (#205693)\n\nTry to address failure in #205221, which results in \u003catomic\u003e not found.\nThis is CMake code copied from other compiler-rt projects using\n\u003catomic\u003e.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: c6a17f1bfd7607229aba71d0de92db95335b8ce7\n" }, { "commit": "17768d32b73335768e2a93ffb55841b8f4c64bde", "tree": "61341ae36ff1c436ecd535d303d89a1eb3c81464", "parents": [ "24cd7325cf57fb6b528a7915d436edc4b50f5f44" ], "author": { "name": "Johannes Doerfert", "email": "jdoerfert.llvm@gmail.com", "time": "Wed Jun 24 22:56:22 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 23:00:54 2026" }, "message": "[Instrumentor] Add runtime examples: [1/N] A flop counter (#205221)\n\nThis adds a instrumentor-tools folder into compiler RT to showcase\nuse cases of the instrumentor. The initial example is a program that,\nvia instrumentation, counts the number of flops performed. Call and\n intrinsic support will follow after #198042.\n\nPartially developed by Claude (AI), tested and verified by me.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: b576ec61928671722c7bf7cf2df42862fd41b683\n" }, { "commit": "24cd7325cf57fb6b528a7915d436edc4b50f5f44", "tree": "971208d9910f0df50c544b22ed6aff8ca98474a1", "parents": [ "8b82075de4f6cf91ec3d648372311b4bd5f7dec0" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Wed Jun 24 19:04:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 19:05:46 2026" }, "message": "[tysan] Guard mallopt() use on SANITIZER_GLIBC instead of SANITIZER_LINUX (#203118)\n\nmallopt() is a glibc extension, not provided by other Linux C libraries\nsuch as musl, but its declaration and its use in\nInitializeInterceptors() are gated on SANITIZER_LINUX. Linking the TySan\nruntime for a musl target therefore fails:\n\n ld.lld: error: undefined symbol: mallopt\n \u003e\u003e\u003e referenced by tysan_interceptors.cpp:215 in\n \u003e\u003e\u003e __tysan::InitializeInterceptors() in libclang_rt.tysan-hexagon.a\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: c46c34284d4a059050fca105ad65acb4f27bfca6\n" }, { "commit": "8b82075de4f6cf91ec3d648372311b4bd5f7dec0", "tree": "c82fdfa9a4e6b0d69bf4f8906a0f366672ca74d7", "parents": [ "b8772d83c8587e3a3b97552676c87ea53b3e8b2a" ], "author": { "name": "Simon Tatham", "email": "simon.tatham@arm.com", "time": "Wed Jun 24 08:34:01 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 08:35:49 2026" }, "message": "[compiler-rt][ARM] Fix underflow handling in new divdf3.S (#204784)\n\nThe code which calculates the \u0027errsign\u0027 parameter to pass to\n`__compiler_rt_dunder` was wrong in two ways. It calculated the value\nwith the wrong sign, and also in the wrong register, r12 rather than r2!\nIn this code\u0027s original context, both of those things made sense (the\n\u0027dunder\u0027 function had a nonstandard ABI). Somehow none of the existing\ntest cases detected the problem.\n\nWe found this bug in a test case downstream that only failed big-endian\n(because that changes which half of the denominator mantissa is left in\nr2 to be accidentally used as errsign). However, the new test cases here\nare designed to detect the failure in both endiannesses.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 099b1f6aeec8058460598106ff0b68fab66b12dc\n" }, { "commit": "b8772d83c8587e3a3b97552676c87ea53b3e8b2a", "tree": "a7e4f1f80bad19d8a77b4f0fbd4f50d3a9f40f1e", "parents": [ "f079fcc71d5c3d0e9b2d9df8aad002d5fd4f3f4e" ], "author": { "name": "Larry Meadows", "email": "Lawrence.Meadows@amd.com", "time": "Mon Jun 22 19:12:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 22 19:15:36 2026" }, "message": "[PGO][HIP] HSA-introspection device profile drain + GPU PGO tests (#203056)\n\n## Summary\n\nFollow-up to #202095 (now landed). #202095\u0027s host-shadow device-profile\ndrain can\nonly collect device counters for kernels that registered a host-side\nshadow via\n`__hipRegisterVar`. Device-linked programs (e.g. RCCL), whose\ninstrumented code\nobjects are linked directly into the device image with no host shadow,\nare never\ndrained.\n\nThis adds a **supplemental, Linux-only HSA-introspection drain** that\nruns after\nthe host-shadow drain: it walks each GPU agent, enumerates only the code\nobjects\nactually resident there, reads each one\u0027s `__llvm_profile_sections`\ntable on the\ndevice, and routes them through the existing `processDeviceOffloadPrf()`\npath so\nthe emitted `.profraw` layout is identical. A content-dedup set keyed on\nthe\n`(data, counters, names)` device-pointer triple ensures a section\nalready drained\nby the host-shadow pass is not drained twice, so the two passes compose\nwithout\ndouble-counting.\n\nIt is purely additive — it does not modify #202095\u0027s host-shadow drain\nor its\nlaunch-tracking. Highlights:\n\n- `compiler-rt/lib/profile/InstrProfilingPlatformROCmHSA.cpp`: HSA\nagent/segment/\nsymbol walk + dedup; record drained bounds after each host-shadow drain;\nlazy\n HSA init (no library constructor, for fork-safety).\n- Because the HSA walk only touches resident code objects, it lets us\navoid the\nhost-shadow drain\u0027s collect-all fallback on Linux. When **no** kernel\nlaunch was\ntracked (program never launches, collects before its first launch, or\nlaunches\nonly via an untracked API), the host-shadow pass is skipped and the HSA\ndrain\ncovers it safely — instead of faulting/hanging reading a non-resident\ndevice on\na multi-GPU host. This also closes the silent-data-loss gap for\nuntracked launch\n APIs (`hipExtLaunchKernel`, cooperative/graph launches).\n- `clang/lib/Driver/ToolChains/Clang.cpp` / `HIPAMD.cpp`: link the\ndevice profile\nruntime on both the new-offload-driver (`LinkerWrapper::ConstructJob`)\nand\ntraditional (`lld`) link paths, guarded by `needsProfileRT` + VFS\nexistence.\n- New GPU/AMDGPU HIP device-PGO lit tests, gated by `hip`/`amdgpu`\nfeatures\n(auto-detected from the toolchain + a visible GPU) so they report\nUNSUPPORTED\nrather than fail when no GPU is present. Plus GPU-free host unit tests\nfor the\ndevice-profile host helpers that run everywhere, including upstream CI.\n\n## Test plan\n\n- 4x gfx90a (`gfx90a:sramecc+:xnack-`), ROCm 7.1.\n- GPU device tests run through standard lit:\n`llvm-lit -sv \u003cbuild\u003e/.../compiler-rt/test/profile/Profile-x86_64` over\nthe\n`GPU/` and `AMDGPU/` subdirectories. The profile `lit.cfg.py`\nauto-detects a\nvisible GPU (`amdgpu-arch`) and the HIP runtime (`libamdhip64`) and\nexposes the\n`hip`/`amdgpu`/`multi-device` features and the `%amdgpu_arch` /\n`%hip_lib_path`\nsubstitutions; both are overridable via `--param amdgpu_arch\u003d… /\nhip_lib_path\u003d…`.\n- **15 device tests passed, 0 failed.** Covers: basic/coverage/pgo-use,\nmultiple-kernels, device-branching, multi-gpu and non-default-device\ndrain,\nearly-collect / no-kernel edges, RDC vs non-RDC\n`__llvm_profile_sections`,\ndedup (host-shadow drains the used device, HSA finds it and dedups), and\n fork-safety (the RCCL parent-no-HIP / kernel-in-forked-child pattern).\n- On a host without a GPU + ROCm/HIP (e.g. upstream CI) those device\ntests report\nUNSUPPORTED instead of failing, and the GPU subdirectories serialize via\na\n size-1 `gpu` lit parallelism group when they do run.\n- GPU-free host unit tests run anywhere the profile suite runs\n(including upstream\n CI): `instrprof-rocm-grow-array.cpp` (the dynamic-array helper) and\n`instrprof-rocm-bounds-dedup.cpp` (the `(data, counters, names)` dedup\ntable\n that backs the \"drain each counter set once\" guarantee).\n- Build is warning-clean and `clang-format` clean.\n\n---------\n\nCo-authored-by: Cursor \u003ccursoragent@cursor.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 55d7f777d958a64ed5ef31be269c3005ddd6c7a6\n" }, { "commit": "f079fcc71d5c3d0e9b2d9df8aad002d5fd4f3f4e", "tree": "af734e7609a729399b3bce44eff498b264598274", "parents": [ "fe03e8f617256a163e2e8c685ae4e0a25cc9ed48" ], "author": { "name": "John Paul Jepko", "email": "john.jepko@ericsson.com", "time": "Mon Jun 22 17:30:11 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 22 17:36:09 2026" }, "message": "[Clang][Sema] Add -Wstringop-overread warning for source buffer overreads (#183004)\n\nThis PR adds a new `-Wstringop-overread` warning that diagnoses calls to\nmemory functions where the specified size exceeds the size of the source\nbuffer, increasing parity with GCC\u0027s `-Wstringop-overread`.\n\nThe warning is emitted when the read size is a compile-time constant\nthat is greater than the size of the source buffer (when known\nstatically).\n\nThis check applies to the following functions:\n- `memcpy`, `memmove`, `mempcpy` (and `__builtin_` / `__builtin___*_chk`\nvariants)\n- `memchr`\n- `memcmp`, `bcmp`\n\nSome of the existing code for `-Wfortify-source` was refactored into a\nhelper class to make its lambdas accessible to other functions.\n\nFixes #83728\n\nAssisted-by: claude-opus-4.6\n\n---------\n\nCo-authored-by: Erich Keane \u003cekeane@nvidia.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 86f2e71cb8d165b59ad31a442b2391e23826133e\n" }, { "commit": "fe03e8f617256a163e2e8c685ae4e0a25cc9ed48", "tree": "5886f8b21baf73777618337c5e99e1bb4290bf35", "parents": [ "a9ea24ac8c7912323f01067b1506b6cf208d5544" ], "author": { "name": "Tamir Duberstein", "email": "tamird@gmail.com", "time": "Sun Jun 21 04:02:19 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 21 04:05:54 2026" }, "message": "[tsan] fit Go/s390x mapping under QEMU (#204503)\n\nQEMU linux-user first tries guest_base\u003d0. In that identity-mapped mode,\nfixed guest mappings use the same host addresses. On an x86-64 host\nwith four-level page tables, the Go/s390x meta shadow starts at\n144 TiB, beyond the 128 TiB userspace limit, and its mmap fails with\nENOMEM during TSan initialization.\n\nMove the meta shadow down by 32 TiB to\n[0x700000000000, 0x780000000000), restoring the 16 TiB gap after the\nshadow and placing all Go/s390x TSan regions below 2^47. Correct the\nmapping comment\u0027s shadow size and ratio.\n\nFailure report and native s390x comparison:\nhttps://github.com/golang/go/issues/67881\n\nQEMU identity guest-base selection:\n\nhttps://github.com/qemu/qemu/blob/v10.2.3/linux-user/elfload.c#L1036-L1042\n\nQEMU guest-to-host address translation:\n\nhttps://github.com/qemu/qemu/blob/v10.2.3/include/user/guest-host.h#L46-L50\n\nQEMU fixed-mapping implementation:\nhttps://github.com/qemu/qemu/blob/v10.2.3/linux-user/mmap.c#L600-L628\n\nx86-64 four-level paging documentation:\nhttps://docs.kernel.org/arch/x86/x86_64/5level-paging.html\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 7376a706150289cf60fb3648e75e909ab6e86f68\n" }, { "commit": "a9ea24ac8c7912323f01067b1506b6cf208d5544", "tree": "3500cc1665b673a9291c2aac4d57fcda2e715a2d", "parents": [ "ea75a440a064dc763eb8f7749ffbcbcf12ddce3a" ], "author": { "name": "Garvit Gupta", "email": "garvgupt@qti.qualcomm.com", "time": "Fri Jun 19 10:20:21 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 19 10:25:53 2026" }, "message": "Revert \"Revert \"[Compiler-rt][test] Fix circular link dependency between builtins and libc\"\" (#204728)\n\nReverts llvm/llvm-project#203152\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: e6daa6810e5ed49104a0f409d51e953905c9752e\n" }, { "commit": "ea75a440a064dc763eb8f7749ffbcbcf12ddce3a", "tree": "0ae0122d37324dcb46d9c2cf00dc3b8c38510ef0", "parents": [ "6796c602b781e4e5bec7f0fe7637ef4a939a8ad9" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Thu Jun 18 23:35:25 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 18 23:40:30 2026" }, "message": "[Fuzzer] Bump rss_limit again (#204701)\n\nThese tests are still somewhat flaky, even after\n3e11b3be3b86d4e72e73e610fb15fe431169f16e, so bump the limit again.\n\nSee\n\nhttps://github.com/llvm/llvm-project/pull/203450#issuecomment-4737237866\nfor info on failures.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: d1c5cf2eac6b09c46adeb5593c11ecc0469b7bb6\n" }, { "commit": "6796c602b781e4e5bec7f0fe7637ef4a939a8ad9", "tree": "1706365e240320c701926578441398d9d9a2f094", "parents": [ "242e3a3fe72644e47bfd432ed03bcd5b38360309" ], "author": { "name": "Sam Elliott", "email": "aelliott@qti.qualcomm.com", "time": "Thu Jun 18 20:13:14 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 18 20:16:13 2026" }, "message": "[RISCV] Implement Zicfiss Extension Bitmask (#201699)\n\nThis implements the proposal here:\nhttps://github.com/riscv-non-isa/riscv-c-api-doc/pull/187\n\nThis was prepared with the assistance of AI.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5471cdb0185c262e8e5fd5219f11095ccf7ddacb\n" }, { "commit": "242e3a3fe72644e47bfd432ed03bcd5b38360309", "tree": "8e6a493cc54c974639f28e266a827a3a2026e82f", "parents": [ "d6d8c5542722c4a9d20e48721dbace26b6e0c6a0" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Thu Jun 18 17:26:17 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 18 17:29:18 2026" }, "message": "[scudo] Remove static error info functions. (#204483)\n\nAndroid was the only platform using the static error info functions. The\nlatest Android no longer uses them, so delete these unused functions.\n\nThese functions were implicitly exporting implementation details and\nmade it difficult to implement the multi-region code. This is the final\nstep to submit the multi-region code and modify how the region code\nworks.\n\nRemove the error info fuzzer since that was only necessary since so many\nvalues were passed in to the function.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 4a86da464e16b8f18d105e620268db3ef808da97\n" }, { "commit": "d6d8c5542722c4a9d20e48721dbace26b6e0c6a0", "tree": "1073a8dfc87c1927a1da1e75276a918fa4e67601", "parents": [ "a7e6fe82976762485f98ca2cefd8d86e8aae52e7" ], "author": { "name": "Andrew Haberlandt", "email": "ahaberlandt@apple.com", "time": "Thu Jun 18 12:35:00 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 18 12:40:43 2026" }, "message": "[compiler-rt] Fix default builtins target _Float16 detection on x86_64/i386 (#204474)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 1343b646697fff896af5c2a0a99135db177ae43d\n" }, { "commit": "a7e6fe82976762485f98ca2cefd8d86e8aae52e7", "tree": "fed14635d78f43433c254e25e7ae596b4dde7917", "parents": [ "0817f0afeee1b37ce8d383b06120ab7308f9243b" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Tue Jun 16 22:55:09 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 16 23:00:10 2026" }, "message": "[scudo] For a realloc that shrinks, retag the extra. (#204031)\n\nWhen MTE is enabled and an allocations is reallocated from a large size\nto a smaller size, zero tag the rest of the allocation. Before this\nchange only a single granule after the new size was zero tagged. This\nadds extra security and use after realloc protection if code would have\ntried to read/write into the old size, past the new size.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5639b33d7567bf648f27418b126f5f6f6292b730\n" }, { "commit": "0817f0afeee1b37ce8d383b06120ab7308f9243b", "tree": "3fec174429852ba344fa86fc95775ec3ed1ee570", "parents": [ "d130f12739bafa2755c9d8054f0bc720efb29edc" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Tue Jun 16 19:33:52 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 16 19:37:47 2026" }, "message": "[scudo] Use the unmap function on MemMap object. (#204001)\n\nThe current call does a unmap(MemMap), but the rest of the code is doing\nMemMap.unmap(XXX), so follow that pattern.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: b2cb99971fb498a7ff458b3a1a4a4edf3c3bafd3\n" }, { "commit": "d130f12739bafa2755c9d8054f0bc720efb29edc", "tree": "ccebcc325f53c82dd089ebf1fb0fcc74c0bc332a", "parents": [ "245c5c44e0ed8f83c339169538c5780faaefbda5" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Tue Jun 16 18:30:17 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 16 18:35:58 2026" }, "message": "[XRay][Hexagon] Use PC-rel addressing for runtime globals in trampoline (#203122)\n\nThe trampolines load the runtime handler globals\n(__xray::XRayPatchedFunction and friends) with absolute\nconstant-extended immediates, which cannot be used in a PIC/PIE link, so\nlinking a default-PIE executable against the xray runtime fails -- and\n-fPIC on user code does not help, the bad relocations are inside the\nruntime archive:\n\n ld.lld: error: relocation R_HEX_32_6_X cannot be used against symbol\n \u0027__xray::XRayPatchedFunction\u0027; recompile with -fPIC\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: c7fbd7e709530a00d42dfb168333139149a0aa98\n" }, { "commit": "245c5c44e0ed8f83c339169538c5780faaefbda5", "tree": "325c139015d0ae167ccbc346c1b4321a5fce8244", "parents": [ "a87056dc9dae783074dc7addd7a6e188ce784570" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Tue Jun 16 18:29:58 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 16 18:35:44 2026" }, "message": "[XRay][Hexagon] Fix immext encoding of high bits in sled patcher (#203129)\n\nencodeConstantExtender() places the high 12 bits of the 26-bit extension\nat the wrong offset (\u003c\u003c16 instead of \u003c\u003c2), dropping them for any\nconstant above ~2^20. The runtime sled patcher then encodes a corrupted\ntrampoline address for PIE executables (load base 0x08000000+), so the\nfirst patched function call jumps to a bogus address and crashes.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 30f73c9cf7a6bd93cdf176912466f6e546f10435\n" }, { "commit": "a87056dc9dae783074dc7addd7a6e188ce784570", "tree": "9e2783db89e9793590032118e3a986106f8c4e0c", "parents": [ "bc97b63a261e142e24df48a613a8ff3254f1c856" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Tue Jun 16 15:47:02 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 16 15:50:23 2026" }, "message": "[XRay][Hexagon] Make XRay sleds register-transparent (#203117)\n\nA sled is inserted post-register-allocation and is treated by the\ncompiler as clobbering nothing, but the sled and trampoline together\npreserve too little. The sled uses r6/r7 as scratch (trampoline address\nand funcid) and the trampoline saves only r0-r5, r31 and p3:0. A\nfunction-exit sled can land in the middle of a loop (a conditional early\nreturn), where loop-carried values occupy r6-r15 and, for a hardware\nloop, SA0/LC0. The handler clobbers them, so most instrumented\ngcc-c-torture programs compute wrong results and abort().\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: fabfc92bcdc2a1dc1164c3a7661c80ef35de06f7\n" }, { "commit": "bc97b63a261e142e24df48a613a8ff3254f1c856", "tree": "88e54b199ba3e96a29fa6beb1fc900b693cf06ba", "parents": [ "c556c9b6513b83b85c47caa639d8672be0f0462f" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Sun Jun 14 18:45:10 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 14 18:50:25 2026" }, "message": "[Fuzzer] Bump max RSS in fork tests (#203688)\n\nThese became flaky on at least one buildbot after enabling the internal\nshell by default:\n1. https://lab.llvm.org/buildbot/#/builders/174/builds/36874\n2. https://lab.llvm.org/buildbot/#/builders/174/builds/36876\n\nTry bumping the max RSS to see if that helps.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3e11b3be3b86d4e72e73e610fb15fe431169f16e\n" }, { "commit": "c556c9b6513b83b85c47caa639d8672be0f0462f", "tree": "99cb14335aee6aca5873b11d6c7f4e81ea54053f", "parents": [ "76b5b7999a9cd7d3fcca6e4d31c894b98fd50988" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Fri Jun 12 20:02:51 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 12 20:05:27 2026" }, "message": "[Fuzzer] Use the internal shell by default (#203450)\n\nThe external shell is going away soon and the internal shell is superior\nin most ways (platform compatibility, performance, debugging). This now\ncauses no test failures and is blocking deprecation of the external\nshell, so switch over.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 327ce1bb0cd56e74bfd79f51463e6c0a81905a25\n" }, { "commit": "76b5b7999a9cd7d3fcca6e4d31c894b98fd50988", "tree": "c63df2f867faa8c8e4a53b7f145157cf1eb64223", "parents": [ "1f414f0765cc26f92525e78dcbf06c639ae2de53" ], "author": { "name": "Jiaqi He", "email": "heturing@gmail.com", "time": "Fri Jun 12 17:55:32 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 12 18:04:17 2026" }, "message": "[RISC-V] Add newer Linux hwprobe extensions detection (#192761)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: e4a63425ba266f6027e7d257c9a79141c26aa8d8\n" }, { "commit": "1f414f0765cc26f92525e78dcbf06c639ae2de53", "tree": "90ed23690a8480d6a765902e1935b3be5d6835e2", "parents": [ "0cc457215a65374947f4d616d2edb1d760851baf" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Fri Jun 12 17:26:16 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 12 17:30:42 2026" }, "message": "[Fuzzer] Make signal tests work with internal shell (#203445)\n\nWrap some commands in bash as it seemed like the least bad option. The\nalternative is to use setsid and run the other commands in the internal\nshell, but them it\u0027s impossible to use wait and we run into reliability\nissues because of that.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: e476b23c5facd1705f06ee60f1e62869c638f395\n" }, { "commit": "0cc457215a65374947f4d616d2edb1d760851baf", "tree": "9ff6ee100e5a5b93ffec564231e830def4bf67d3", "parents": [ "f2db5499c1eadb3bf0b7c8f75c18bbbff6e2f1c3" ], "author": { "name": "Yaxun (Sam) Liu", "email": "yaxun.liu@amd.com", "time": "Fri Jun 12 13:24:05 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 12 13:25:50 2026" }, "message": "[PGO][HIP] Fix HIP device profile collection and sections emission (#202095)\n\nSeveral related HIP device-PGO fixes:\n\nWindows device collection. HIP rejects a hipMemcpy that reads past the\nbounds\nof a symbol registered with __hipRegisterVar, but device\ndata/counters/names\nlive in merged linker sections. Register a separate shadow for each\ndevice\ndata, counters, and names symbol and copy each one by its exact\nhipGetSymbolSize\nsize; this also lets static TUs with several kernels keep all their\nprofile\ndata. Open the device profile file in binary mode and pass the device\nnames to\nthe correct lprofWriteDataImpl arguments so llvm-profdata can read the\nraw\nprofile. Open the versioned amdhip64_7.dll first, falling back to\namdhip64.dll.\n\nPer-TU sections struct. Clang CodeGen emitted the\n__llvm_profile_sections_\u003cCUID\u003e\nstruct (and its section start/stop references) for any profiling-enabled\ndevice\nTU. A TU with no instrumented device functions then referenced sections\nnothing\npopulates, so the RDC device link failed under --no-undefined (and\nduplicated\n__llvm_prf_nm before per-CUID naming). Move the struct emission from\nCGCUDANV\ninto the InstrProfiling pass, which emits it only when the TU has\nprofile data;\nclang emits only the per-TU names-postfix marker, also making names\nunique per\nTU so RDC builds do not clash.\n\nDynamic-module interceptors. The hipModuleLoad* interceptors live in a\nconstructor-only object in clang_rt.profile_rocm that nothing\nreferences, so the\nlinker drops it and dynamic-module programs collect no device profile.\nWhen\nlinking clang_rt.profile_rocm, emit a force-link reference (-u on ELF,\n-include: on COFF); the constructor self-skips when the program does not\nuse\nhipModuleLoad.\n\nMulti-device profile collection. On Linux, static profile collection\nused to\ntry reading profile data from every visible HIP device. This could fault\nwhen a\ndevice was visible but had not launched the instrumented kernel. Track\nHIP\ndevices that successfully launch kernels, and skip unused devices during\nstatic\nprofile collection. If tracking is not available, keep the old\ncollect-all\nbehavior.\n\nDepends on #201607 (reland HIP offload PGO compiler support and link the\ndevice-profile runtime); that PR must land first.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 7c0a3a52cf967da9c41d009fb92453b272d0d04a\n" }, { "commit": "f2db5499c1eadb3bf0b7c8f75c18bbbff6e2f1c3", "tree": "2a979922af7223c7bb2df573bc8ad9e7cc32133c", "parents": [ "18ea7eceea38dd6cf6eb44b92f5707eeb673cab0" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Fri Jun 12 03:28:16 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 12 03:30:29 2026" }, "message": "[Fuzzer] Make two tests compatible with the internal shell. (#203448)\n\n1. Remove redundant parntheses that broke the internal shell\u0027s parsing\nlogic.\n2. Use env when specifying environment variables.\n3. Rewrite a bash one-line loop in python.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 2a6cfc5643053117b3baa301b91afaf9fc5edede\n" }, { "commit": "18ea7eceea38dd6cf6eb44b92f5707eeb673cab0", "tree": "278fed039718a934d5b886816283e1cdcc80aeb6", "parents": [ "acdef5091d9dd1c780c5a8198b13e54642502a74" ], "author": { "name": "David Tenty", "email": "daltenty@ibm.com", "time": "Thu Jun 11 02:36:19 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 11 02:40:57 2026" }, "message": "Revert \"[Compiler-rt][test] Fix circular link dependency between builtins and libc\" (#203152)\n\nReverts llvm/llvm-project#199482 due to failures when it\u0027s used on\nplatforms with non-ELF linkers. The patch needs additional guards, but\nit\u0027s not immediately clear which platform linkers support the required\noptions.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 7588c957cf4f686ce4ced4b1df717695be295bf0\n" }, { "commit": "acdef5091d9dd1c780c5a8198b13e54642502a74", "tree": "f623a7207c73ad0fc69d5ae35e67fc1bed0ea8d2", "parents": [ "f70e8d392d2371e9d91c56bb0808c14b92e52992" ], "author": { "name": "Justin T. Gibbs", "email": "gibbs@scsiguy.com", "time": "Wed Jun 10 16:54:12 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 10 16:55:15 2026" }, "message": "[asan] Fix asan_new_delete.cpp C++ header resolution under -nostdinc++ (#202816)\n\nThe COMPILER_RT_ASAN_ENABLE_EXCEPTIONS gate in asan/CMakeLists.txt\nenables -fexceptions on the C++ slice (RTAsan_cxx /\nRTAsan_dynamic_cxx — asan_new_delete.cpp) and tries to expose C++\nstandard headers to that TU (for forthcoming std::bad_alloc support)\nby stripping -nostdinc++ from its cflags. The strip works for native\nstandalone builds but is wrong for cross builds: the host C++ headers\naren\u0027t valid for the target.\n\nSplit the C++-slice flag handling into two paths inside the existing\nEXCEPTIONS gate:\n\n * In-tree libc++ available (TARGET cxx-headers OR HAVE_LIBCXX):\n keep -nostdinc++, append ${COMPILER_RT_CXX_CFLAGS} (a generator\n expression that expands to \"-isystem \u003cprepared cxx-headers dir\u003e\"),\n and add cxx-headers to DEPS so the header tree is staged before\n the compile. Mirrors the pattern in orc / fuzzer / memprof /\n tsan / xray.\n\n * No in-tree libc++ (typical standalone build): drop -nostdinc++\nfor the C++ slice TU only so the host toolchain supplies C++ headers.\n\nReject COMPILER_RT_CXX_LIBRARY\u003dnone +\nCOMPILER_RT_ASAN_ENABLE_EXCEPTIONS\u003dON\nat configure time: that combination opts out of any C++ stdlib while\nasking for an exception-enabled C++ slice, silently leaking host headers\ninto a `none`-mode build.\n\nAssisted by: Claude Opus 4.7\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: c5c33597bfdba185a6f8d04c0330e386a2e52abf\n" }, { "commit": "f70e8d392d2371e9d91c56bb0808c14b92e52992", "tree": "a82322461524b221e1cf3d4cb28c8a71c271ca1a", "parents": [ "5358b42cbbc6f1492c78461c119d4dd4372e4173" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Wed Jun 10 16:48:03 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 10 16:50:05 2026" }, "message": "lsan: Document that ptrace_scope needs to be disabled.\n\nAs far as I\u0027m aware, this is the reason why\nhttp://45.33.8.238/linux/201640/step_9.txt is failing.\n\nReviewers: vitalybuka\n\nPull Request: https://github.com/llvm/llvm-project/pull/202838\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 06cc1baa0e0eced548df56e3f1fdb7df10e4fd86\n" }, { "commit": "5358b42cbbc6f1492c78461c119d4dd4372e4173", "tree": "231f3130de8942f73154d94e946b71c5278373ab", "parents": [ "a3d1856fb52a9ab50d77fcb048b98e029d086f14" ], "author": { "name": "Simon Tatham", "email": "simon.tatham@arm.com", "time": "Wed Jun 10 08:27:08 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 10 08:28:38 2026" }, "message": "[compiler-rt][ARM] Optimized integer -\u003e FP conversions (#179928)\n\nThis commit adds a total of 8 new functions, all converting an integer\nto a floating-point number, varying in 3 independent choices:\n\n* input integer size (32-bit or 64-bit)\n* input integer type (signed or unsigned)\n* output float format (32-bit or 64-bit)\n\nThe two conversions of 64-bit integer to 32-bit float live in the same\nsource file, to save code size, since that conversion is one of the more\ncomplicated ones and the two functions can share most of their code,\nwith only a few instructions differing at the start to handle negative\nnumbers (or not).\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 93e03fc2666e10b4e32091bc27d35709f49c9443\n" }, { "commit": "a3d1856fb52a9ab50d77fcb048b98e029d086f14", "tree": "291df5a924b4c8f97c3bc5a0ae3cc96fba777838", "parents": [ "db25b263c5ac0256e912c2708356f5b7c071ed14" ], "author": { "name": "Yaxun (Sam) Liu", "email": "yaxun.liu@amd.com", "time": "Tue Jun 09 20:57:07 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 21:00:14 2026" }, "message": "Build ROCm profile runtime by default (#202718)\n\nThe ROCm profile runtime should get regular build coverage. Keeping it\noff by\ndefault lets it regress unless a build enables it explicitly.\n\nMake the option default to on for normal Linux and Windows builds, where\nthe\nruntime is supported. The target still has the existing dependency\nchecks, and\nnormal profile links keep using the base profile runtime unless the\ndriver\nselects the ROCm archive for HIP profiling.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 4ed07e26fc00fbfef25b7e9889eace5514a795f0\n" }, { "commit": "db25b263c5ac0256e912c2708356f5b7c071ed14", "tree": "431ba566b95271e7bdd74168c9c59b2bf05599d2", "parents": [ "53db5eeb819376c73c9a8eecbe7ee56b4fac2354" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Jun 09 18:29:05 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 18:30:28 2026" }, "message": "[compiler-rt] Fix undefined AMDGPU variable not set for builtins (#202733)\n\nSummary:\ncompiler-rt is weird and has multiple config files, make sure this is\nset for both libs and builtins.\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3c0e69e655a28a0061404b5f296a07abf037e5be\n" }, { "commit": "53db5eeb819376c73c9a8eecbe7ee56b4fac2354", "tree": "3cb6a93b7253a5d7d6d449d802e4282a47b3cf00", "parents": [ "e111978fa010060792a5a972635057f511f402e3" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Jun 09 17:43:44 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 17:45:59 2026" }, "message": "compiler-rt: Consolidate regex checks for amdgpu targets (#202281)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 14b2935dba9eab5a52aa2a8c806e29102fbafcf6\n" }, { "commit": "e111978fa010060792a5a972635057f511f402e3", "tree": "da92698dedb983120cbdf7bcdef0d546af0bb9d7", "parents": [ "11bdbc5e4ac14526a5383fe23f5f2cc38c356d4f" ], "author": { "name": "gulfemsavrun", "email": "gulfem@google.com", "time": "Tue Jun 09 17:09:32 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 17:11:59 2026" }, "message": "[asan] Make exceptions for asan_new_delete.cpp conditional (#202436)\n\nIntroduce the COMPILER_RT_ASAN_ENABLE_EXCEPTIONS CMake\noption to control whether the ASan C++ runtime\n(asan_new_delete.cpp) is compiled with exception\nsupport.\n\nThis fixes build failures on platforms with noexcept\ntoolchains (like Fuchsia\u0027s noexcept variant) where\nASan was compiled with exceptions but linked against a\nnoexcept libc++abi, resulting in undefined symbol\nerrors for __cxa_begin_catch and __gxx_personality_v0.\n\nThe option defaults to ON to preserve the behavior of\n#200719, but automatically defaults to OFF if\nLIBCXX_ENABLE_EXCEPTIONS or LIBCXXABI_ENABLE_EXCEPTIONS\nis set to OFF in the same runtimes build.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 95bd483fbb6c7b94dd8fa9d8577d633aa02da6b8\n" }, { "commit": "11bdbc5e4ac14526a5383fe23f5f2cc38c356d4f", "tree": "871c3679207a7d1414ab1579af9d89af85a31f19", "parents": [ "b78374f3219da5cfa343e280be83acae6266f359" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Jun 09 16:56:13 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 17:00:35 2026" }, "message": "compiler-rt: Suppress -g error for gpu builds (#202230)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: f4ee842902bf1fb29a956b6a36b17c547407cdbf\n" }, { "commit": "b78374f3219da5cfa343e280be83acae6266f359", "tree": "b1e0f36fde1491a8ea4dccfe4f0a1064cab45647", "parents": [ "49bbbb08fa975c862b9fb878a57407f5338bcb85" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Jun 09 13:49:22 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 13:50:35 2026" }, "message": "[compiler-rt] Fix misspelled variable name in GPU caches (#202604)\n\nSummary:\nThis likely didn\u0027t do anything, but it\u0027s best for things to be spelled\nright.\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 67e3b31450012b939980fc91075d0ab0b035f284\n" }, { "commit": "49bbbb08fa975c862b9fb878a57407f5338bcb85", "tree": "10771238c1835945a09349c753287100785bfad2", "parents": [ "b824a366462eb6fd44cbfc9a630aeff35ba003d5" ], "author": { "name": "Xiaomeng Zhang", "email": "zhangxiaomeng@hygon.cn", "time": "Tue Jun 09 10:45:52 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 10:48:00 2026" }, "message": "[X86] Hygon Processors Initial enablement (#187622)\n\nThis patch adds initial support for several Hygon architectures.\n\nThe Hygon architectures include:\n\n- c86-4g-m4\n- c86-4g-m6\n- c86-4g-m7\n\nThis patch includes:\n\n- Added Hygon architectures CPU targets recognition in Clang and LLVM\n- Added Hygon architectures to target parser and host CPU detection\n- Updated compiler-rt CPU model detection for Hygon architectures\n- Added Hygon architectures to various optimizer tests\n- Added scheduler models for Hygon architectures CPU targets\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 00e3e6f4cbbba60ddfcd648269dd5f2e8d799cbc\n" }, { "commit": "b824a366462eb6fd44cbfc9a630aeff35ba003d5", "tree": "01f39c7207bb2a9e6dc9d4c06b6f19a428b84faa", "parents": [ "6b925cd18ac0b924703d9adb4980f411ae0ae133" ], "author": { "name": "Garvit Gupta", "email": "garvgupt@qti.qualcomm.com", "time": "Tue Jun 09 08:42:27 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 08:45:50 2026" }, "message": "[Compiler-rt][test] Fix circular link dependency between builtins and libc (#199482)\n\nCurrently, the link order is `libclang_rt.builtins.a -lc -lm`. Builtins\nare scanned first after which symbols like `abort` are unresolved\nreferences that are resolved through libc.a. However, resolving the\nreferences to these symbols further lead to undefined references to\n`_aeabi_uldivmod` etc. that can only resolved through builtins.\nReversing the order also wont fix the issue because `libc.a` introduces\n`__aeabi_uldivmod` which is resolved by builtins but it introduces\n`abort` which can only be resolved libc.a.\n\nThis patch fixes this by wrapping the archives in a linker group\n(--start-group/--end-group), which instructs the linker to rescan all\narchives in the group until no new symbols can be resolved.\n\nThis error is exposed only when bfd like linkers are used.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 8b4902300521d4a0980d9d35210c02b405f0df86\n" }, { "commit": "6b925cd18ac0b924703d9adb4980f411ae0ae133", "tree": "419d6036610491d9282031e00cbbd4014aff8ccf", "parents": [ "3e6c74e31fb0b09dfec238982ac6dbfbe80be2bd" ], "author": { "name": "Jakob Koschel", "email": "jakobkoschel@google.com", "time": "Tue Jun 09 07:39:34 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 07:40:46 2026" }, "message": "[SafeStack] Introduce public interface header and deprecate builtins (#198292)\n\nIntroduce `sanitizer/safestack_interface.h` to expose SafeStack utility\nfunctions as a public interface, similar to other sanitizers.\n\nExposed functions:\n* `__safestack_get_unsafe_stack_ptr`\n* `__safestack_get_unsafe_stack_bottom`\n* `__safestack_get_unsafe_stack_top`\n\nDeprecate the existing `__builtin___get_unsafe_stack_*` builtins and\nemit warnings suggesting the new `__safestack_` equivalents.\n\nRuntime compatibility aliases `__get_unsafe_stack_*` are maintained.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 69215c5e4f03253d78218bf5ac55f81b90118082\n" }, { "commit": "3e6c74e31fb0b09dfec238982ac6dbfbe80be2bd", "tree": "830951905003b535d7ae19aa33bb1e81423ba20b", "parents": [ "940408af6881744c723082a4471784dd65988dc4" ], "author": { "name": "Brian Cain", "email": "brian.cain@oss.qualcomm.com", "time": "Tue Jun 09 01:33:54 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 09 01:35:28 2026" }, "message": "[rtsan][clang] Add Hexagon support for RTSan (#200313)\n\nEnable RTSan for the Hexagon architecture.\n\n* Add Hexagon to ALL_RTSAN_SUPPORTED_ARCH in cmake\n* Add a clang driver test for hexagon-unknown-linux-musl\n* Guarding a static_assert(sizeof(unsigned long) \u003e\u003d sizeof(off_t)) with\nSANITIZER_WORDSIZE \u003e\u003d 64, since off_t syscall args are split into two\nregs.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 9ab6d33bc4d30bce9de62e2d96e0f3427ae0eddf\n" }, { "commit": "940408af6881744c723082a4471784dd65988dc4", "tree": "d3655fba2fd66df2a99c9fdf8636b132dabd4806", "parents": [ "584f93f164abe81f168aaeb4cea4fff12b59bcf5" ], "author": { "name": "David Justo", "email": "david.justo.1996@gmail.com", "time": "Mon Jun 08 20:09:46 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 08 20:10:31 2026" }, "message": "[ASan] add pattern \u0027cmp BYTE PTR [rdx], XX\u0027 to win instruction decoder (#202407)\n\n**Context:** The ASan instruction decoder in `interception_win.cpp` has\na manual case-based list for each instruction pattern we expect to see\nin function prologues.\n\nToday, we have an instruction decoder for `cmp BYTE PTR [rcx], XX`, but\nwe do not have the equivalent for `cmp BYTE PTR [rdx], XX`. In recent\nbuilds of Windows, that latter is now seen in `ucrtbase!strstr`.\n\n**This PR** adds the missing case.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3ca5b7dc4f50f5e1af51f25c2575c8fa509c24b7\n" }, { "commit": "584f93f164abe81f168aaeb4cea4fff12b59bcf5", "tree": "dea499a90eace8a5107709fe0d77651221b719d7", "parents": [ "cbe8159640b48cd7abf541813710ba1eadb45705" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Mon Jun 08 19:09:36 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 08 19:11:34 2026" }, "message": "[scudo] Remove locks from getTSDAndLockSlow (#201960)\n\nChange the name of getTSDAndLockSlow to getTSDLockSlow and remove all\nlocks in favor of using atomic variables. Continue to lock when changing\nthe number of shared TSDs.\n\nThis change will remove thread contention and avoid blocking between\nthreads when this function is called.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3136d62938e8cb8ac3ca735d7db741a0c0d11521\n" }, { "commit": "cbe8159640b48cd7abf541813710ba1eadb45705", "tree": "4ac06c73d596b50991ceb01a5af701b93ea58e29", "parents": [ "ca4c0f8db0509db1bab0bb957dbf20eee6b7b2a0" ], "author": { "name": "Justin T. Gibbs", "email": "gibbs@scsiguy.com", "time": "Mon Jun 08 18:02:27 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 08 18:05:49 2026" }, "message": "[asan] Enable exceptions for asan_new_delete.cpp translation unit (#200719)\n\nGroundwork for #196388.\n\nPreparation for a follow-up change that lets the throwing operator new\nwrappers in asan_new_delete.cpp throw std::bad_alloc on OOM. To support\nthat, asan_new_delete.cpp must be compiled with -fexceptions and must be\nable to include \u003cnew\u003e (i.e. without -nostdinc++). The rest of the ASan\nruntime stays compiled with -fno-exceptions and -nostdinc++.\n\nNOTE: Comments in CMakeLists.txt are written assuming the follow-on\n changes to use exceptions land, thus avoiding the need to update\n this file again.\n\nBuild-system changes:\n\n* Define ASAN_CXX_CFLAGS as ASAN_CFLAGS minus -fno-exceptions and\n-nostdinc++ plus -fexceptions. Define ASAN_DYNAMIC_CXX_CFLAGS likewise\nrelative to ASAN_DYNAMIC_CFLAGS. RTTI stays disabled; libstdc++ /\nlibc++abi supply the bad_alloc typeinfo.\n\n* Static build: switch RTAsan_cxx (which already contains only\n${ASAN_CXX_SOURCES} \u003d asan_new_delete.cpp) from ASAN_CFLAGS to\nASAN_CXX_CFLAGS. RTAsan (the C slice) is unchanged.\n\n* Dynamic build: split the previous single RTAsan_dynamic object library\ninto two — RTAsan_dynamic keeps ${ASAN_SOURCES} (no change in flags) and\na new RTAsan_dynamic_cxx carries ${ASAN_CXX_SOURCES} with\nASAN_DYNAMIC_CXX_CFLAGS. Both shared clang_rt.asan link rules pull in\nRTAsan_dynamic_cxx alongside RTAsan_dynamic so the linked .so retains\nthe same set of object files; only the compile flags for\nasan_new_delete.cpp change.\n\nNFC at runtime — the asan_new_delete.cpp source as of this commit does\nnot yet use exceptions or include \u003cnew\u003e, so the new flags are inert. The\nfollow-up commit makes use of them.\n\nAssisted by: Claude Opus 4.7\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 8b45994309c19752dd369e86487461f93c00532b\n" }, { "commit": "ca4c0f8db0509db1bab0bb957dbf20eee6b7b2a0", "tree": "72abd9aedc09d44385ebd1ddeec0e3e5ff7a4fcd", "parents": [ "ba0d6cb047ad6e83ac91e2e30b0052cebd2147b3" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Mon Jun 08 16:55:27 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 08 17:00:21 2026" }, "message": "LowerTypeTests: Mark CFI jump table sections as eligible for relaxation.\n\nUse !elf_section_properties metadata to set the type and entry size to\nthe correct values, and set the preferred alignment to the entry size\nto enable last jump table entry placement.\n\nReviewers: vitalybuka, MaskRay\n\nReviewed By: vitalybuka, MaskRay\n\nPull Request: https://github.com/llvm/llvm-project/pull/149261\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 081568eb664543d62babfe882fa8bbfa6a353ce0\n" }, { "commit": "ba0d6cb047ad6e83ac91e2e30b0052cebd2147b3", "tree": "95510eaadda8783d5e4ad8c6b6edd161c35209ba", "parents": [ "265393e6a03798e2fa0678f9eb29e6d0c6d54929" ], "author": { "name": "Fangrui Song", "email": "i@maskray.me", "time": "Mon Jun 08 06:19:14 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 08 06:20:34 2026" }, "message": "[sanitizer_common] DenseMap: replace tombstone deletion with TAOCP 6.4 Algorithm R (#202231)\n\nsanitizer_dense_map.h is a fork of llvm/ADT/DenseMap.h, which uses\nquadratic probing with lazy deletion: an erased entry becomes a\ntombstone, a third bucket state alongside empty and live that every\nfind/insert must inspect.\n\nPort the upstream #200595 and getTombstoneKey() removal.\n\nAided by Claude Opus 4.8\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: d1cbede06860efefdef36de926004edb13fa3228\n" }, { "commit": "265393e6a03798e2fa0678f9eb29e6d0c6d54929", "tree": "a73ddd35e20af7b4661417f3a3b2fde1906168f7", "parents": [ "87087aa47ee704e4b650d5f49277bf3db8c7390a" ], "author": { "name": "Matt Turner", "email": "mattst88@gmail.com", "time": "Sun Jun 07 05:51:54 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 07 05:55:57 2026" }, "message": "[ASan] Improve qemu-alpha shadow mapping (#201861)\n\nWith a 1T fixed shadow offset the usable app memory is split between\nLowMem (0-1T) and HighMem (1.5T-4T). This works on real Alpha hardware\nwhere all addresses stay within TASK_SIZE (4T). However, under\nqemu-alpha user mode mmap(NULL) returns addresses from the host x86-64\naddress space (~127T), outside both regions, causing AddrIsInMem() CHECK\nfailures in PoisonShadow.\n\nSwitch to a fixed shadow offset of 0x70000000000 (7 TiB). TASK_SIZE is\nwell below the shadow offset so HighMem is empty: kHighMemBeg \u003d\nMEM_TO_SHADOW(kHighMemEnd) + 1 \u003e kHighMemEnd. All app memory fits in\nLowMem [0, 7T), a simpler layout with no HighMem split. On qemu-alpha,\n-R 0x80000000000 constrains guest mappings to [0, 8T), keeping them\nwithin LowMem.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 825b3c71ecdc2f1eee9a5d46180532d3b9a12a67\n" }, { "commit": "87087aa47ee704e4b650d5f49277bf3db8c7390a", "tree": "eacaf9229b66e082e75d1ef2fc6e256d02995725", "parents": [ "f508792acf075548730b4b626365f024b3994328" ], "author": { "name": "Matt Turner", "email": "mattst88@gmail.com", "time": "Sat Jun 06 18:34:31 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Jun 06 18:36:11 2026" }, "message": "[ASan] Skip high-shadow and gap setup when HighMem region is empty (#202037)\n\nOn targets where the shadow offset sits above all addressable user\nmemory (e.g. Alpha with ASAN_SHADOW_OFFSET\u003d0x70000000000 and a 42-bit\nuser VAS), kHighMemBeg is set above kHighMemEnd so the HighMem region is\nempty. Since MEM_TO_SHADOW is monotonically increasing, kHighMemBeg \u003e\nkHighMemEnd implies kHighShadowBeg \u003e kHighShadowEnd. Calling\nReserveShadowMemoryRange(kHighShadowBeg, kHighShadowEnd) passes size \u003d\nkHighShadowEnd - kHighShadowBeg + 1, which underflows to a large\nnegative value, and mmap() fails with ENOMEM.\n\nProtectGap is also skipped: there is no meaningful shadow gap between\nLowShadow and an empty HighShadow.\n\nGuard both operations on kHighMemBeg \u003c\u003d kHighMemEnd.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 8e1afdada5d09cccb18b3ff17dac04e30645b93e\n" }, { "commit": "f508792acf075548730b4b626365f024b3994328", "tree": "fac461454411919436738cd91bee2da0334733b8", "parents": [ "fbac051b680fc20032520e7272ba3b3f76bb696d" ], "author": { "name": "Daniel Thornburgh", "email": "dthorn@google.com", "time": "Sat Jun 06 00:28:51 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Jun 06 00:30:35 2026" }, "message": "[compiler-rt] Add missing test stdlib.h include (#201972)\n\nFixes test after libc++ PR #195509 which drops transitive includes.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 7ef07bca9b05b4a3435aa857dbd417c2aec40c22\n" }, { "commit": "fbac051b680fc20032520e7272ba3b3f76bb696d", "tree": "2a087d8fe88af8d8af9860ae8dc6122f6f7808a0", "parents": [ "3e5f27bbe216da29c1fe5f780ab7e5dc2669fee5" ], "author": { "name": "Daniel Thornburgh", "email": "dthorn@google.com", "time": "Fri Jun 05 22:40:19 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 22:45:54 2026" }, "message": "[compiler-rt] Add missing cstdlib include to signal_line.cpp test (#201955)\n\nFixes test after libc++ PR #195509 which drops transitive includes.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 1dae5dee50e73d18ef42027bc3bd853a9ae740ce\n" }, { "commit": "3e5f27bbe216da29c1fe5f780ab7e5dc2669fee5", "tree": "a3eb71e54fbb0574626ccb02a433a79681c82a08", "parents": [ "e2c01e4b969ed5b2859c78ce79f79e282cc2bdec" ], "author": { "name": "Reid Kleckner", "email": "rkleckner@nvidia.com", "time": "Fri Jun 05 22:24:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 22:30:51 2026" }, "message": "[docs] Rename 20 key LLVM docs to Markdown (#201464)\n\nThis may break the docs build temporarily, but there will be a follow-up\nPR that fixes it immediately after. The rename is being committed first\nto aid review and source code archaelogy.\n\nTracking issue: #201242\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 638f3ae61de291f85b711f6e3597fae6edd29986\n" }, { "commit": "e2c01e4b969ed5b2859c78ce79f79e282cc2bdec", "tree": "d0b0684ad8f461d25d13db37370a0d6f601cc680", "parents": [ "6cb1ddbbac2d52f63e3ae5e1a76f4239615e3c32" ], "author": { "name": "Daniel Thornburgh", "email": "dthorn@google.com", "time": "Fri Jun 05 21:52:40 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 21:55:50 2026" }, "message": "[compiler-rt] Add stdlib include to OutOfProcessFuzzTarget.cpp (#201944)\n\nFixes test after libc++ PR #195509 which drops transitive includes.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: e9a4dd0b325c8d171e68662fbea554a86757a79f\n" }, { "commit": "6cb1ddbbac2d52f63e3ae5e1a76f4239615e3c32", "tree": "1b6308310c2c20122289e81396d76e2f83212747", "parents": [ "7c920f7bb6a8ccfcc31419acfa43a4ac0344e3ac" ], "author": { "name": "Florian Mayer", "email": "fmayer@google.com", "time": "Fri Jun 05 21:37:08 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 21:40:56 2026" }, "message": "[compiler-rt] [sanitizers] Add includes for symbolizer (#201929)\n\nAfter ce5b2e876494cb95f02d9f915081e2b8781e74d1 this stopped building.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 3b89e77151214ec62546f7437b4d5e07b1c2ce1d\n" }, { "commit": "7c920f7bb6a8ccfcc31419acfa43a4ac0344e3ac", "tree": "30f480f9dcf944640870147ee1b242370a61e10b", "parents": [ "b7b5c1c594f1a3b9964cd4d67f72174b4fde8a8a" ], "author": { "name": "Daniel Thornburgh", "email": "dthorn@google.com", "time": "Fri Jun 05 19:33:45 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 19:35:40 2026" }, "message": "[orc] Add missing \u003citerator\u003e include (#201916)\n\nFixes build after libc++ PR #195509 which drops transitive includes.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: ae48a36859750575a9d9166b5ebd604656741ad1\n" }, { "commit": "b7b5c1c594f1a3b9964cd4d67f72174b4fde8a8a", "tree": "70d08d099dc79b3824b3a3500c28ed85e49d6bb6", "parents": [ "9e2658de6f69e1b3d36bc01a6cbb2a3131da199e" ], "author": { "name": "Nikolas Klauser", "email": "nikolasklauser@berlin.de", "time": "Fri Jun 05 07:45:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 07:50:51 2026" }, "message": "[libc++] Drop transitive includes by default (#195509)\n\nThis patch removes the unused transitive includes by default.\n`_LIBCPP_KEEP_TRANSITIVE_INCLUDES_LLVM23` can be defined to keep the\ntransitive includes around for an easier transition. The macro will be\nremoved in LLVM 24.\n\nThis patch implements\nhttps://discourse.llvm.org/t/rfc-remove-unused-transitive-includes-from-the-libc-headers/90157\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: ce5b2e876494cb95f02d9f915081e2b8781e74d1\n" }, { "commit": "9e2658de6f69e1b3d36bc01a6cbb2a3131da199e", "tree": "60aa38ffecce19ee77d87bd2796c29d809543875", "parents": [ "7d789f434b6d25d91d7092e2358c385b81e22eff" ], "author": { "name": "Brad Smith", "email": "brad@comstyle.com", "time": "Fri Jun 05 04:32:48 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 04:35:33 2026" }, "message": "[LLVM][compiler-rt][RISCV] Refactor RISC-V CPU features (#199438)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 0b0a345d18626801395b6ca957af0a1be5aecc1e\n" }, { "commit": "7d789f434b6d25d91d7092e2358c385b81e22eff", "tree": "7b79b6808f9388b411920fe2d4b7682aa9369eb9", "parents": [ "60a6314227289781b8124827ab0d413952011782" ], "author": { "name": "Yaxun (Sam) Liu", "email": "yaxun.liu@amd.com", "time": "Fri Jun 05 00:24:01 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 00:25:37 2026" }, "message": "Reland HIP offload PGO runtime support as a separate opt-in library (#201606)\n\nThis mostly relands the compiler-rt part of #177665 (approved and\nmerged, then reverted in #201416). The first commit restores it as\nmerged.\n\nIt was reverted because of a Windows problem: the ROCm runtime needs the\nsanitizer interception library, which is built /MD on Windows, so\nputting it in clang_rt.profile forced that library to /MD and broke\nusers linking it with the static CRT (/MT).\n\nThe second commit fixes this by building the ROCm support as a separate,\nopt-in library clang_rt.profile_rocm, a /MD superset of\nclang_rt.profile. The base library is left unchanged (/MT, no ROCm). The\ndriver links clang_rt.profile_rocm first, so it resolves all profile\nsymbols and the base library stays inert.\n\nclang_rt.profile_rocm is off by default. The compiler-side change and\ndriver wiring are in a separate PR.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: ef574da13acec7d30de1e949d8a2c7c6f7ff531f\n" }, { "commit": "60a6314227289781b8124827ab0d413952011782", "tree": "40a77778f0d2af5c9d2ed4158d77f24a480870bb", "parents": [ "a9bdc5ff3a92d811c198955747f1ab9d131fb308" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Thu Jun 04 19:06:10 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jun 04 19:11:30 2026" }, "message": "[scudo] Log if randomness degrades. (#201482)\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 19703323e0c75c63214fcc29272c118326b95ecc\n" }, { "commit": "a9bdc5ff3a92d811c198955747f1ab9d131fb308", "tree": "186958fa1c2c9565c22ab0ab769dafb2812a6fdf", "parents": [ "85b0ac81b41c2f2fc05cdd395353d8fb93878861" ], "author": { "name": "Miguel A. Arroyo", "email": "miguel.arroyo@rockstargames.com", "time": "Wed Jun 03 20:36:37 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 03 20:40:28 2026" }, "message": "[compiler-rt][cmake][msvc] Install PDBs for shared libraries (#201224)\n\nFollowing up on similar PRs in the past to address PDBs not being copied\nfor install targets:\n* https://github.com/llvm/llvm-project/pull/126961\n* https://github.com/llvm/llvm-project/pull/126680\n* https://github.com/llvm/llvm-project/pull/120683\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: fb649b41b0db8cd33a5be613345be845a41d0d98\n" }, { "commit": "85b0ac81b41c2f2fc05cdd395353d8fb93878861", "tree": "bc238f2ceb2e62ed3a1359e9e4fce5f8b7b9bce9", "parents": [ "5d18abbff30b3713c209e9b49959a01363cb37e3" ], "author": { "name": "George Burgess IV", "email": "george.burgess.iv@gmail.com", "time": "Wed Jun 03 18:55:02 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 03 19:01:33 2026" }, "message": "[compiler-rt] fix out-of-bounds access in scudo test (#201390)\n\nWhen `I \u003d 0`, we can\u0027t do `Array[I-1]`.\n\nCaught by libcxx hardening.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5708e7fd9eca368b1b519c13a58360a15a263e4f\n" }, { "commit": "5d18abbff30b3713c209e9b49959a01363cb37e3", "tree": "74a218dd2b226d1ce26090c139f9a40b83bfb837", "parents": [ "45a5d9099951a5665d673912ed2198daa3800b85" ], "author": { "name": "Hans Wennborg", "email": "hans@hanshq.net", "time": "Wed Jun 03 18:05:47 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 03 18:10:58 2026" }, "message": "Revert \"[PGO][AMDGPU] Add basic HIP offload PGO support (#177665)\" (#201416)\n\nThis broke profiling builds on Windows by switching the profile library\nto link against the dynamic CRT; see discussion on the PR.\n\nThere were already a number of issues reported and fixed after this PR.\nRather than piling on the fixes (and this one may need some work),\nrevert back to green for now to let the project recover.\n\nThis reverts commit 5db13643f4b7038db0ca304d9f8900122502935c.\n\nAdditionally, this reverts the followup PRs in\n635e120fb87304924508a7a204574727e3c37363,\n2766733764f4bdf8399d48c0225e9c64bdfa95f7,\n4c33844b4b560f24b2a3a0cc689d73510cb01ad5, and\n5eca8b67ff3c5f371141d54b6e4544a0ebe77fdb:\n\n\"[PGO][HIP] Stop pulling ROCm.o into every PGO host link (#200101)\"\n\"[compiler-rt][profile] Add COMPILER_RT_BUILD_PROFILE_ROCM option\n(#200127)\"\n\"[PGO][HIP] Skip ROCm interceptor in profile-only compiler-rt builds\n(#200111)\"\n\"[PGO][HIP] Fix profile-only Windows link by gating ROCm interceptor\nmacro (#200859)\"\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 6cfa1a01a0f737ed1d54963810057fb1bd67a274\n" }, { "commit": "45a5d9099951a5665d673912ed2198daa3800b85", "tree": "14f1d02de7dd9dd90bb09ec58491dea29b99d189", "parents": [ "0bbf712d62680b1cebb7f411ab93fa5fe1e87646" ], "author": { "name": "Amit Kumar Pandey", "email": "pandey.kumaramit2023@gmail.com", "time": "Tue Jun 02 09:52:49 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 02 09:55:44 2026" }, "message": "[compiler-rt][sanitizer_common] Generalize CheckNoDeepBind as OnDlOpen (#200748)\n\nRename the dlopen pre-check hook to OnDlOpen so platform-specific dlopen\nhandling can be extended beyond the RTLD_DEEPBIND guard on Linux.\nWindows and macOS keep no-op implementations. All dlopen interceptors\ncall the shared hook.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: bae4566a422cba33fb41a072320b36fb51cb67ee\n" }, { "commit": "0bbf712d62680b1cebb7f411ab93fa5fe1e87646", "tree": "e55ba105e4dc201576350d587ebd9aac34220053", "parents": [ "a4abe1ba3af2db298c0194f8767b97afb5c21b94" ], "author": { "name": "Christopher Ferris", "email": "cferris1000@users.noreply.github.com", "time": "Tue Jun 02 04:49:18 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jun 02 04:50:31 2026" }, "message": "[scudo] Return nullptr if a remap fails on linux. (#200537)\n\nAdd a check if a fixed address mmap doesn\u0027t return the expected address.\n\nAllow a remap call to fail if the mmap fails and returns a nullptr to\nthe caller.\n\nFix a place where if remap fails in the secondary, it didn\u0027t do\nanything. Now it will unmap the original entry on failure.\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: adcd285fc4047759e4014d6568d80bb64589aded\n" }, { "commit": "a4abe1ba3af2db298c0194f8767b97afb5c21b94", "tree": "49611c37302209926fb86edf5ca4bb5355511bce", "parents": [ "9889c7e9db4a970fe4e8e16abf86ee602626e02a" ], "author": { "name": "Wael Yehia", "email": "wmyehia2001@yahoo.com", "time": "Mon Jun 01 18:45:34 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 01 18:50:50 2026" }, "message": "[AIX] Set the ifunc constructor\u0027s priority to 100, to run before any user code (#200893)\n\nCo-authored-by: Wael Yehia \u003cwyehia@ca.ibm.com\u003e\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 9433fd7666ca6e0e38c05e3dc911b0685ffa0ab3\n" }, { "commit": "9889c7e9db4a970fe4e8e16abf86ee602626e02a", "tree": "253471636af3d81752d08d1aaddd2134959298d9", "parents": [ "b899dc508d1da3eb172ba8a65ae6d1176b0358cf" ], "author": { "name": "Yaxun (Sam) Liu", "email": "yaxun.liu@amd.com", "time": "Mon Jun 01 16:55:15 2026" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 01 17:00:24 2026" }, "message": "[PGO][HIP] Fix profile-only Windows link by gating ROCm interceptor macro (#200859)\n\nPR #200111 stops compiling InstrProfilingPlatformROCm.cpp (which defines\nthe\nHIP GPU helper __llvm_profile_hip_collect_device_data) in profile-only\nbuilds.\nBut the compile define -DCOMPILER_RT_BUILD_PROFILE_ROCM\u003d1 was still\nadded\nwhenever the COMPILER_RT_BUILD_PROFILE_ROCM option was on (the default),\nso\nInstrProfilingFile.c still referenced the helper from\n__llvm_profile_write_file\neven though it was never built.\n\nOn ELF the declaration is weak, so the undefined symbol folds to null\nand the\naddress-guarded call is skipped. COFF/Windows has no such fallback:\n\n error LNK2019: unresolved external symbol\n __llvm_profile_hip_collect_device_data referenced in function\n __llvm_profile_write_file\n\nAdd the define only when PROFILE_HAS_HIP_INTERCEPTOR is true, i.e. the\nsame\ncondition that keeps InstrProfilingPlatformROCm.cpp in the archive, so\nthe\nmacro is defined iff the helper is actually compiled in.\n\nReported by zmodem:\nhttps://github.com/llvm/llvm-project/pull/200111#issuecomment-4593893230\n\nNOKEYCHECK\u003dTrue\nGitOrigin-RevId: 5eca8b67ff3c5f371141d54b6e4544a0ebe77fdb\n" } ], "next": "b899dc508d1da3eb172ba8a65ae6d1176b0358cf" }