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

KDE Git Services - Bulk Change <[email protected]>
Newsgroups gmane.comp.kde.cvs
Message-ID <[email protected]>
Git repository change summary for qt/clang/llvm
Pushed by mirror-service into branch 'upstream/users/shiltian/fix-crash-invalid-sdmw-operand-dasm'.
Changed from 0000000000000000000000000000000000000000 to 113befc2cdc76abd9cb0ef7fa74918e92d18df2c
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 b5f3fed2801d71ed04218975f2a9eceb4146f087 by GitHub (on behalf of Florian Hahn) on 09/08/2026 at 17:30..
[VPlan] Move structural loop-region verification to verifier. (#215079)

VPlan::getVectorLoopRegion() finds the vector loop region by following
the last successor of each block, starting at the plan's entry. Verify
that plans have the shape this relies on: the entry must have no
predecessors, the region must be the only top-level loop region, and
each block on the path from the entry to the region must have a single
predecessor which has that block as its last successor.

This replaces the EXPENSIVE_CHECKS-only assert by checking in the
verifier, allowing better handling for VPlans violating the expectations
in unit tests.

Fixes an assertion with expensive checks after
https://github.com/llvm/llvm-project/pull/199437:
https://lab.llvm.org/buildbot/#/builders/187/builds/22485.
https://invent.kde.org/qt/clang/llvm/-/commit/b5f3fed2801d71ed04218975f2a9eceb4146f087

Git commit 3d8868224e00d2f3821380c7df985c61fd5dfab3 by GitHub (on behalf of Chris Apple) on 09/08/2026 at 17:36..
[sanitizer-common][rtsan] Mark unsupported tests (#213404)

From #213170, turning that PR into separate commits as suggested.

To enable rtsan in sanitizer common, first disable the tests that are
not trivially fixable (or will remain permanently unsupported).


Quoting from that other review, assuming all of these are merged and
rtsan is enabled for sanitizer-common
---
Darwin:
New tests that run and pass: 249
UNSUPPORTED: 26
Total: 275
UNSUPPORTED percentage: 9.5%

Linux is about the same, few additional unsupported:
New tests that run and pass: 249
UNSUPPORTED: 40
Total: 289
UNSUPPORTED percentage 13.8%

Again, some of these UNSUPPORTED are temporary, and we'll go back
through and clean them up (or at least discuss suport) so I expect this
percentage to go down over time
https://invent.kde.org/qt/clang/llvm/-/commit/3d8868224e00d2f3821380c7df985c61fd5dfab3

Git commit 08cb7d93f7e88de47c50f41ae08514f17c624f7a by GitHub (on behalf of Fangrui Song) on 09/08/2026 at 17:59..
[DomTree] Store immediate dominator as DFS number. NFC (#215059)

Replace the local variable `IDoms` and the pointer `InfoRec::IDom` with
a DFS number member.

Since `attachNewSubtree` and `reattachExistingSubtree` walk NumToNode in
DFS order, they can index IDom directly. Replace an always-false `if
(DT.getNode(W))` condition with an assert.
https://invent.kde.org/qt/clang/llvm/-/commit/08cb7d93f7e88de47c50f41ae08514f17c624f7a

Git commit 51de808337a1c98761bdcee40524589392c9e7f9 by GitHub (on behalf of Fangrui Song) on 09/08/2026 at 19:18..
[BFI] Detect irreducible SCC from CycleInfo instead of by failing. NFC (#214941)

BFI computes block frequencies with the Wu-Larus algorithm, extended (in
2014) to irreducible SCCs by modelling each one as a loop with multiple
headers.

Irreducible SCCs are detected by failing: mass distribution runs until
`addToDist` hits a retreating local edge to a non-header, then aborts,
packages the SCCs it found, and reruns.

CycleInfo already reports reducibility, so mark the innermost enclosing
loop (or the top level) while initializeLoops walks the cycle forest,
and package the marked regions before distributing any mass.  The abort
paths and both undo sites become asserts.

Aided by Claude Opus 5
https://invent.kde.org/qt/clang/llvm/-/commit/51de808337a1c98761bdcee40524589392c9e7f9

Git commit 113befc2cdc76abd9cb0ef7fa74918e92d18df2c by Shilei Tian on 09/08/2026 at 20:10..
[AMDGPU][MC] Fix a crash when invalid SDWA encoding is used

Fixes #215006.
https://invent.kde.org/qt/clang/llvm/-/commit/113befc2cdc76abd9cb0ef7fa74918e92d18df2c
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.