[LAA] Don't check for free in evaluatePtrAddRecAtMaxBTCWillNotWrap() (#202341)
The fact that an object of a certain size existed at the the location at
some point is sufficient to prove the desired nowrap fact. Whether the
object still exists there doesn't matter in this context.
[libc++] Simplify some meta programming in <variant> (#201538)
Specifically, this avoids `__dependent_type` and `__type_identity`
instantiations, reducing compile times a bit.
[AMDGPU] Track VALU instructions separately for WMMA coexecution hazards
WMMA coexecution hazards can only be resolved by VALU instructions, not
S_NOPs. Track VALU/WMMA instructions separately so the scheduler can
accurately determine stall cycles.
[AMDGPU] Drop docs for invalid load-release and store-acquire operations (#202338)
The LangRef says "release and acq_rel orderings are not valid on load
instructions" [for loads](https://llvm.org/docs/LangRef.html#load-instruction)
and "acquire and acq_rel orderings aren't valid on store instructions"
[for stores](https://llvm.org/docs/LangRef.html#store-instruction).
Providing them in textual IR is diagnosed with an error.
Therefore, we should not define semantics for these invalid constructs.
Part of LCOMPILER-2273.
Reland [RISCV] Check SP-relative offset in needsFrameBaseReg when FP offset overflows (#202499)
When a frame pointer is present, needsFrameBaseReg previously only
checked the FP-relative offset to decide if a virtual base register was
needed. If the worst-case FP offset exceeded the 12-bit immediate range,
a base register was always materialized, even when the SP-relative
offset would fit.
Since getFrameIndexReference can now select SP over FP when the offset
fits in the compressed instruction immediate range, also check the
SP-relative offset before deciding a base register is needed. This
avoids unnecessary base register materialization and results in some
code size savings.
Reland of b5d577d3faef.
[clang][bytecode] Refactor backtrace param printing (#202325)
Instead of `classify()`ing the parameter types here, just use the saved
`PrimType` in the `ParamDescriptor`.
This also fixed an oddity with `interp::Context`: `Pointer::toRValue()`
takes an `interp::Context`, but some call sites passed an `ASTContext`
instead, which worked because of the implicit constructor. Make it
explicit.
Fix missing TLF action on Closing -> Initial transition
When a Down event occurs in the Closing state, the FSM transitions
to Initial without a TLF action. Previously, the PPPoE layer was not
explicitly stopped during this transition, leaving the PPP phase
stuck in SPPP_PHASE_ESTABLISH.
As a result, if an Open event occurred afterward, the PPPoE layer
would wait for the PPPOE_SLOW_RETRY delay instead of reconnecting
immediately.
Added an additional TLF action to reset the phase and ensure
immediate reconnection upon the next Open event.
interfaces: parse ifconfig output despite exit error in legacy_interfaces_details()
PR: https://github.com/opnsense/src/issues/284
In this version print the error number and stderr output. We'd like to know
which part of ifconfig fails and don't want to taint the parser output. If
it's empty it's empty anyway. But maybe in the commit we merge we should just
ditch everything to /dev/null as we do mute the error in most cases anyway.
The resulting array is empty worst case which is just what it is.
Suggested by: @e-alfred
compiler-rt: Consolidate regex checks for amdgpu targets
In the future the triple naming scheme will change, and this
will help avoid repeating the same longer regex in all of these
places.
Co-Authored-By: Claude Sonnet 4.5 <noreply at anthropic.com>
[AIX][libc++] Mark offset_range test UNSUPPORTED on 32-bit AIX due to 32-bit off_t (#200861)
Marking the test as UNSUPPORTED and combined the comment since both
platforms have the same fundamental limitation with `32-bit off_t`.
---------
Co-authored-by: himadhith <himadhith.v at ibm.com>
interfaces: parse ifconfig output despite exit error in legacy_interfaces_details()
PR: https://github.com/opnsense/src/issues/284
In this version print the error number and stderr output. We'd like to know
which part of ifconfig fails and don't want to taint the parser output. If
it's empty it's empty anyway.
Suggested by: @e-alfred
