[qt/clang/llvm-project]: Summary of bulk changes made

Git repository change summary for qt/clang/llvm-project
Pushed by mirror-service into branch 'upstream/release/23.x'.
Changed from ce6af707aac85db6e176ad0faa3bfe608e1a68f5 to c7280675cd04a97c958bcf46db8042e33ab35fc9
Acknowledgement was received that this change introduces only existing code that has been pushed to another public open source repository.

This change contains the following new commits:

Git commit f54355e08d91e1f4c87337505936e9299a9ffb99 by llvmbot (on behalf of 陈子昂) on 16/07/2026 at 08:00..
[LoopUnroll] Invalidate SCEV after full unrolling (#208874)

Full unrolling may leave stale ScalarEvolution loop and disposition
caches around the loop pass boundary. A later IndVars pass can then use
stale exit count information and incorrectly fold reachable exits.

Invalidate SCEV loop and disposition caches after successful full
unrolling before preserving `ScalarEvolutionAnalysis.`

Fixes #207744.

(cherry picked from commit 42c2980c7c632d7a45776d3e2c51627100f6e763)
https://invent.kde.org/qt/clang/llvm-project/-/commit/f54355e08d91e1f4c87337505936e9299a9ffb99

Git commit a2520c9f136c302b05b501c05222e0f605897b32 by Douglas Yung (on behalf of Karthika Devi C) on 16/07/2026 at 16:26..
[Polly] Skip vectorize.enable for FP loops with dist=1 dependences (#205756)

When -polly-annotate-metadata-vectorize is active, Polly marks its
generated loops with llvm.loop.vectorize.enable=true. This is harmful
for loops with a loop-carried dependence of distance 1 that involve
floating-point operations: the Loop Vectorizer reorders FP operations
(e.g. scalar reduction like q = factor*q), producing results that differ
from the sequential scalar reference and causing correctness failures.

Two changes are made:

1. IslAst.cpp: add PollyVectorizeMetadata to the PerformParallelTest
gate so that dependence-distance computation is performed whenever
-polly-annotate-metadata-vectorize is passed, not only when
-polly-parallel or a vectorizer is active.

2. IslNodeBuilder.cpp / LoopGenerators.cpp: when a loop has a dist=1
dependence involving FP operations, omit the vectorize.enable annotation
entirely. This lets the Loop Vectorizer apply its own cost model and
legality checks decide.

(cherry picked from commit df1838f5d961f9587ca4f90d992bef81cc6155db)
https://invent.kde.org/qt/clang/llvm-project/-/commit/a2520c9f136c302b05b501c05222e0f605897b32

Git commit 23780df286c2e5914014d002206f1bfde7d0f1b9 by Douglas Yung (on behalf of Karthika Devi C) on 16/07/2026 at 16:27..
[Polly] Fix codegen assertions to account for DefinedBehaviorContext (#209188)

DeLICM may produce new read access relations whose domain is restricted
to the DefinedBehaviorContext (e.g., only valid when a parameter ensures
no UB). The validation in setNewAccessRelation already accounts for
this, but the debug assertions in createNewAccesses and
generateScalarLoads did not, causing false assertion failures during
code generation.

Intersect the checked domains with getBestKnownDefinedBehaviorContext()
to match the contract that DeLICM relies on.

Fixes #205732

(cherry picked from commit ea612545644d1e3f238ac4cd053758f569f9bf3b)
https://invent.kde.org/qt/clang/llvm-project/-/commit/23780df286c2e5914014d002206f1bfde7d0f1b9

Git commit 88c69e55f3da48bcebdb425f54431f80ce457c8d by Douglas Yung (on behalf of Amy Kwan) on 16/07/2026 at 16:27..
[Object][GOFF] Recognize RLD and LEN records (#207118)

This patch adds explicit cases for RT_RLD and RT_LEN records when
parsing GOFF objects. These record types are not handled yet, but
recognizing them allows us to diagnose them and avoids teating them as
unexpected records.

(cherry picked from commit c40401dd8d28e88519433531c041b0b7e5ada8a9)
https://invent.kde.org/qt/clang/llvm-project/-/commit/88c69e55f3da48bcebdb425f54431f80ce457c8d

Git commit c7280675cd04a97c958bcf46db8042e33ab35fc9 by Douglas Yung (on behalf of Erich Keane) on 16/07/2026 at 16:28..
Fix references to complete types in attribute references (#209537)

This is a regression from #197215.

Attributes are not REALLY in the body of a function (though the name of
said function is... awkwardly inaccurate at best), but still need to pay
attention to the completeness of their references. As a result, we
weren't marking the expression as invalid, but were also trying to
evaluate it.

This patch fixes this in 2 ways. First, we re-add the
CXXThisTypeOverride check, but except constant substitution, since that
has some additional 'this' behavior from #197215. x

Secondly, we also make the constant evaluator give up on incomplete
types when handling an L value member. This stops us from trying to
evaluate the value if it is incomplete during template instantiation,
when the type is incomplete. We don't diagnose, since it is still
potentially a constant expression, but isn't currently one.

Fixes: #199527
(cherry picked from commit 0d93f2aec0b4847761bd8458b7847ce37b85cf03)
https://invent.kde.org/qt/clang/llvm-project/-/commit/c7280675cd04a97c958bcf46db8042e33ab35fc9