Summary: 697 instances, 623 unique

Text	Count
// It is kind complex128 but not type complex128. TODO: We can handle this unsafely.	2
// TODO: this includes references of indexed symbols from other packages,	1
// TODO: Move into funcinfo.	1
// FIXME: why does this need to be set to zero?	1
// TODO: consider trimming leading zeros.	1
// TODO: make cgo not depend on $GOARCH?	1
// It is kind uintptr but not type uintptr. TODO: We can handle this unsafely.	2
// TODO: remove this restriction. See issue 47631.	1
// TODO: What is the latency cost of this? Tune this value.	1
// TODO: Anything for tagProfile_DropFrames?	1
// TODO: pass safe to packEface so we don't need to copy if safe==true?	2
// TODO: It would be nice if ExecError was more broken down, but	1
C_S3216CON = C_32S16CON // TODO: these should be treated differently (e.g xoris vs addis)	1
// TODO: do we need to update named values here?	1
// TODO: hashedrefs unused/unsupported for now	1
// TODO handle case where l is irreducible.	1
// TODO: stop walking the iterated domininance frontier when	1
// FIXME: is there a way we could avoid materializing the	1
// TODO: double-check this is handled correctly when	1
// TODO: this could be quadratic if lots of variables are live across lots of	2
// TODO: if large prefixes of sep are matching	1
// TODO: check that any declared content-length matches, like	1
// TODO: We could probably make this faster by	1
// TODO: This is only used for debug printing. Maybe export config.registers?	1
// TODO: Check for error.	1
// TODO: if a single value is in multiple registers, spill one of them	1
// (TODO: or can they? Needs a test.)	1
// TODO: should we catch panics from the called method?	1
// FIXME: use an aux sym or a relocation here instead of a	1
// TODO: switch to FPXX after be sure that no odd-number-fpr is used.	1
// TODO: These are dregs left by Makefile-based builds.	1
// TODO: Report inlined functions.	1
// TODO: See issue 5279. The printing of commands needs a complete redo.	1
// TODO: This is somewhat overkill, we really only need it	1
// TODO: there is a Clever Hack that allows pre-generation of a small-ish number of the slices	1
// TODO: if b.Kind == BlockPlain, start the analysis in the subsequent block to find	1
// TODO: alignment on AIX	1
// TODO: Order these to minimize jump distances?	1
MOVL    (DI), BX    // TODO: This turns bad writes into bad reads.	1
// TODO: implement	1
// TODO take into account actual LLC size. E. g. glibc uses LLC size/2.	1
// TODO: handle more cases; shortcircuit optimizations turn out to be reasonably high impact	1
// TODO: Don't busy wait. This loop should really only	1
// TODO: jump across functions needs reloc	1
// TODO: setup TLS.	3
// TODO: For OpStoreWB and the buffered write barrier,	1
// TODO: Add the static type we're converting from as well.	1
case ssa.OpCopy: // TODO: use MOVQreg for reg->reg copies instead of OpCopy?	1
// TODO: We currently use the 2-byte instruction TESTB AX, (reg).	2
// TODO: reuse overflow buckets instead of using new ones, if there	4
mu sync.Mutex // TODO: maybe switch to RWMutex	1
// R(1)<<...   Ugly check. TODO: Rethink how we handle ARM register shifts to be	1
// TODO: adjust buffer writing sizes based on server config, frame size updates from peer, etc	1
// TODO: Look up nelems from sizeclass and remove this field if it	1
// TODO: use LastIndexByte once the bootstrap compiler is >= Go 1.5.	1
// TODO: return error if server tries to RST_STREAM an idle stream	1
if r.Add()&0xff000000 == 0xeb000000 { // BL // TODO: using r.Add here is bad (issue 19811)	1
// TODO: is this redundant? How could we be in malloc	1
// FIXME: add caching or reuse reloc slice.	1
// TODO: enable this padding for all allocations, not just	1
// TODO: For big endian, these slices should be in reverse byte order,	1
// TODO: For big endian, use the same index values as found	1
var cipherSuites = []*cipherSuite{ // TODO: replace with a map, since the order doesn't matter.	1
// TODO figure out exactly what's unused, don't spill it. Make liveness fine-grained, also.	1
// TODO: if this path winds up being taken frequently, it	2
// TODO: Note that this fails if nil is shadowed.	1
// TODO: use os.ReadDir	1
MOVD    (R2), R0    // TODO: This turns bad writes into bad reads.	1
// TODO: Could use git cat-file --batch.	1
// FIXME: would be better to do away with this and have a better way	1
// TODO: we should probably do the same for unions. Unions can't live	1
// TODO: Prefer registers with already spilled Values?	1
// TODO above.	1
// TODO: Unify this with gentraceback and CallersFrames.	1
// TODO: Since traceAllocBlock is now go:notinheap, this isn't necessary.	1
// TODO: improve this part	1
// TODO add dynimport ordinal support.	2
// TODO: also do this if we've written the headers but not	1
// TODO: opt: use a perfect hash function instead of a global map.	1
// TODO: Since prove now derives transitive relations, it	1
// TODO: other transformations in spec or practice?	1
// TODO: we would like to depend only on the types2 computation of structural type,	1
// TODO: have some other way to detect the difference between	1
// TODO: Consider making two passes where the first	1
case ssa.OpCopy: // TODO: use MOVLreg for reg->reg copies instead of OpCopy?	1
// TODO: more of "8.1.2.2 Connection-Specific Header Fields"	1
// TODO: Use this as the return type of sysAlloc, persistentAlloc, etc?	1
MOVD    $0, 0(R15)              // clear back chain pointer (TODO can we give it real back trace information?)	1
// TODO: SetMaxDynamicTableSize, SetMaxDynamicTableSizeLimit on	1
// It is kind float64 but not type float64. TODO: We can handle this unsafely.	2
// TODO if this remains identical to ClosureAuxCall above after new ABI is done, should deduplicate.	1
// TODO: Set m.Defn properly if we support inlining range statement in the future.	1
// FIXME: at the moment all.bash does not pass when I leave out	1
// TODO replace all uses of CtzXX with TrailingZerosXX; they are the same.	1
// TODO: Eliminate more nil checks.	1
aux := ssa.StaticAuxCall(ir.Syms.Deferproc, s.f.ABIDefault.ABIAnalyzeTypes(nil, ACArgs, ACResults)) // TODO paramResultInfo for DeferProc	1
// TODO: handle some more cases, like where	1
// TODO: record and return stderr.	1
// TODO: fix one day?	2
// TODO: read from sysconf(_SC_IOV_MAX)? The Linux default is	1
// FIXME: This doesn't work if an archive has two members with the same	1
// TODO: if b contains only OpCopy or OpNot related to b.Controls,	1
// TODO: instantiations at least know the shape of the instantiated	1
// FIXME: find a better mechanism for this	1
t.Align = 1 // TODO: should probably base this on field alignment.	1
// TODO: The .swig and .swigcxx files can use	1
// TODO: it's possible (in FOR loops, in particular) for statement boundaries for the same	1
// TODO: Test and delete this condition.	1
// TODO: set types directly here instead of using typecheck.Stmt	1
// TODO: For now, simply ignore the field. In the near	1
// TODO: Would be nice if go/doc did this for us.	1
// TODO: For big endian, bytes don't need to be reversed.	1
// TODO: remove more Connection-specific header fields here, in addition	1
// TODO: unsupported encoding	1
// TODO: detect cycles before calling Marshal which loops infinitely on	1
// TODO: add support for using the kind in the non-typeparam case.	1
{name: "late opt", fn: opt, required: true}, // TODO: split required rules and optimizing rules	1
var cipherSuitesTLS13 = []*cipherSuiteTLS13{ // TODO: replace with a map.	1
// TODO: map err to more interesting error codes, once the	1
// TODO: Provide access to TODOs and NOTEs as well (very noisy so off by default)?	1
// TODO: Is this check needed?	1
// TODO: Handle types defined within function.	1
// FIXME: it would be nice if the compiler attached an aux symbol	1
r2.URL = cloneURL(r.URL) // legacy behavior; TODO: try to remove. Issue 23544	1
// TODO: Impose limits on command output size.	1
// TODO: maybe we should introduce these extra phis?	1
// g0 TODO: Don't modify g here since R30 is nonvolatile	1
g.sigcode1 = 0 // TODO: emulate si_addr	1
// TODO: all.bash runs compilers in parallel. Need to synchronize logging somehow?	1
// It is kind bool but not type bool. TODO: We can handle this unsafely.	2
// TODO: remove this string search and be more like the Windows	1
if p.As == AB || p.As == obj.ARET || p.As == AERET { /* TODO: other unconditional operations */	1
// TODO: Perhaps allow comparison between signed and unsigned integers.	1
// TODO: could negate condition(s) to merge controls.	1
// TODO: consider handling cases where:	1
// reading data (see possible TODO at top of	1
// TODO: Perfect for go:nosplitrec since we can't have a safe point	1
idx := ir.NewConstExpr(constant.MakeUint64(uint64(i)), dict) // TODO: what to set orig to?	1
// TODO: when split information is recorded in export data, insert checks only on backedges that can be reached on a split-call-free path.	1
// TODO: Performance here is really sensitive to the	1
// TODO: configurable?	1
// TODO: do clients send GOAWAY too? maybe? Just Close:	1
// TODO: let getReadBuf be configurable, and use a less memory-pinning	1
// See TODO in regalloc.go.	1
v *Value // pre-regalloc value held in this register (TODO: can we use ID here?)	1
// TODO: are there other cases (e.g., wrapper functions) that need marking?	1
MOVD    R0, 0(R1)    // clear back chain pointer (TODO can we give it real back trace information?)	1
// need to change.  TODO: This should be cleaned up.	1
<!-- TODO The rule below needs to be generalized to interface elements.	1
// TODO: take into account Likely property for forward/backward	1
// TODO: Not really a problem; could change API without effect but	2
// TODO: Worth doing better? Probably not, because we're here only	1
// TODO: lazy binding?	1
// TODO: Make OpAddr use AuxInt as well as Aux.	1
// TODO: detect defaults for len/cap also.	1
MOVW    (R20), R1    // TODO: This turns bad writes into bad reads.	1
// TODO: map out.MaxPathLen to 0 if it has the -1 default value? (Issue 19285)	1
// TODO: configure the architecture	1
// TODO: if the variable is dead at c, skip it.	1
// TODO: Anonymous fields.	1
// TODO: Consider caching this in the running G.	1
// TODO: ignoring lang in sv[1] for now. If anybody needs it we'll	1
nx := ir.NewSelectorExpr(base.Pos, ir.OXDOT, np, f.Sym) // TODO: fields from other packages?	1
// Need to emit to be safe (however, see TODO above).	2
// TODO: make this part of cmd/compile/internal/ssa somehow?	1
// TODO: We could clear out buffers just before mark if this	1
// TODO: MXCSR register?	1
// TODO: implement for clobberdead experiment. Nop is ok for now.	8
// TODO: Should have original compilation	1
// It is kind int but not type int. TODO: We can handle this unsafely.	2
// TODO: enforce types of Aux in this switch (like auxString does below)	1
// TODO all the calls to this are overkill; only need to do this for slots that are not present in the merge.	1
// TODO: never happens because pointers are directIface?	1
// TODO: emptyRest is hard to use here, as we start iterating	1
// TODO when registers really arrive, must also decompose anything split across two registers or registers and memory.	1
// TODO return error if found .cormeta	1
// TODO: Since traceBuf is now go:notinheap, this isn't necessary.	1
list *mSpanList // For debugging. TODO: Remove.	1
// TODO finish intrinsifying 386, deadcode the assembly, remove build tags, merge w/ intrinsics_common	1
// TODO: Should we allow else-if in with and range?	1
// TODO: this parsing is somewhat lame and not exact.	1
// TODO: It would be nicer if we could	1
// TODO: make this smarter, using a list of packages + substring matches.	1
ABI1           *abi.ABIConfig // "ABIInternal" under development // TODO change comment when this becomes current	1
// TODO: This file contains the special casing rules for Turkish and Azeri only.	1
// TODO: IsPrint is a local implementation of unicode.IsPrint, verified by the tests	1
// TODO: find a better place for this logic.	1
// TODO: pass safe boolean from valueInterface so	2
// TODO: For big endian, the index should be 31 not 0.	1
// TODO: unify this makefield and the global one above.	1
// TODO: Handle 32-bit comparisons.	1
// TODO: other unrolling idioms	1
// TODO: return an error? golang.org/issue/14048	1
atomic.Xadd(&sched.npidle, -1) // TODO: fast atomic	1
// TODO: don't close a cc if it was just added to the pool	1
// TODO maybe cache this in a map if profiling recommends.	1
// TODO: This padding is only performed when the race detector	1
// TODO: maybe take size instead of type?	1
case token.EQL, token.NEQ: // TODO: more cases. This is enough for the cftype fix.	1
// TODO results should not be lowered until this phase.	1
http2defaultMaxStreams      = 250 // TODO: make this 100 as the GFE seems to?	1
rewriteNewPhis(h, c, f, defsForUses, newphis, dfPhiTargets, sdom) // TODO: convert to explicit stack from recursion.	1
// TODO: maybe the compiler could set the alignment for all	1
// TODO: care about exact syntax of chunk extensions? We're	1
// TODO: Implement this	6
// TODO: care? unlike http/1, it won't mess up our framing, so it's	1
// TODO: Test and delete these conditions.	1
// TODO what if control is statement boundary? Too late here.	1
// It is kind uint but not type uint. TODO: We can handle this unsafely.	2
// TODO should also optimize offsets from SB?	1
// TODO: maybe this could be	1
Alignment:     4, // TODO: just for arm5?	1
// TODO: pick a less arbitrary value? this is a bit under	1
// TODO: need to work FP comparisons into block jumps	1
// TODO: consider supporting target.Is(err). This would allow	1
// FIXME: at the moment attributes are stored in a linked list in a	1
// TODO: specify which of the caller and callee is responsible for IP	1
// FIXME: It should be forbidden to have R_ADDR from a	1
{name: "opt", fn: opt, required: true},               // NB: some generic rules know the name of the opt pass. TODO: split required rules and optimizing rules	1
// TODO if too slow, whole program iteration can be replaced w/ slices of appropriate values, accumulated in first loop here.	1
// TODO: Implement solaris (see https://golang.org/issue/8106)	1
// TODO: Set environment to get English error messages.	1
name   string // TODO: would this be better as offset into str table?	1
// TODO need to check bounds?	1
WB                 bool         "help:\"enable write barrier\"" // TODO: remove	1
// TODO: the start PC is returned only when cache is nil.	1
type SymbolType int8 // TODO: after genasmsym is gone, maybe rename to plan9typeChar or something	1
// TODO: do not emit sfcall if operation can be optimized to constant in later	1
p.errorf("TODO: implement three-operand instructions for this architecture")	1
// TODO:	2
// TODO: it would be safer to only do this if	1
// TODO: When the old backend is gone, reconsider this decision.	1
// TODO(bradfitz,brainman): use ctx more. See TODO below.	1
// TODO: get rid of some of these temporaries.	1
// TODO: remove this once we've matched the parser resolution exactly.	1
// It is kind uint32 but not type uint32. TODO: We can handle this unsafely.	2
// TODO: It would be much better if we didn't	1
// TODO: pick apart structs, do them piecemeal too	1
#TODO fix quoting for qualified variable names	1
// TODO: Prologue and epilogue are perhaps better handled as pseudo-ops for the assembler,	1
// TODO: Should we filter out nils in the fast	1
// TODO: deprecate these; use internal/cpu directly.	1
// TODO: is this still worth it with new memmove impl?	1
v := c.NewValue0I(currentRoot.Pos, ssa.OpPhi, typ, int64(n)) // TODO: line number right?	1
// FIXME: might be worth looking replacing this map with a function	1
// TODO: remove this helper function once http2 can use build	1
// TODO: Improve this part. At least the size of endRegs of the predecessor also has	1
// TODO: could we also get into this state if	1
if !exportFormat { // TODO maybe only if FmtShort, same below	1
timer := time.NewTimer(http2prefaceTimeout) // TODO: configurable on *Server?	1
// TODO: Remove.	1
// FIXME: remove this O(n) loop	1
} // TODO and if this fails?	1
MOVV    R0, 2(R0) // TODO: TD	1
// TODO: This would be cleaner if sysReserve could be	1
w.string(string(b)) // TODO: More efficient encoding.	2
// TODO: Mark it "already zeroed"	1
// TODO: The second branch is probably predict-not-taken since it is for FP unordered	1
// It is kind string but not type string. TODO: We can handle this unsafely.	2
// FIXME: these are now unions, so maybe need to change definitions?	1
// TODO: this could be more efficient allocation-wise:	1
// TODO: Is elem guaranteed valid? If so we could avoid this check.	1
hasIrreducible bool // TODO current treatment of irreducible loops is very flaky, if accurate loops are needed, must punt at function level.	1
// TODO: Split attr and element name part filters so we can recognize known attributes.	1
//    TODO: Maybe someday we label them all with NOFRAME and get rid of this heuristic.	1
// TODO: add more here, after collecting data on the common	4
// TODO: remove indirect function call when golang.org/issue/30548 is fixed	2
// TODO: We could do more by being unsafe. Add a -unsafe flag?	2
// TODO: test on non-zero extraBytes param.	1
// TODO: fix the above behavior and cleanup this exception.	1
// TODO: improve the prediction of the likely predecessor. The following	1
// only valid until the next Scan/Close. But the TODO is that	1
// TODO: build and vcs information is not embedded for executables in GOROOT.	1
// TODO: remove r2 from zsyscall_aix_$GOARCH.go	1
// TODO: export/import types and addrtaken marks along with inlined bodies,	1
// TODO: MOVB might be ok if we only ever use R15B.	1
// TODO: could make this a bit faster by having a helper	1
// TODO: track which type of frame & with which flags was sent	1
// TODO: share slots among equivalent types. We would need to	1
// TODO: use singleflight for dialing and addConnCalls?	1
// TODO: kind of a hack - allows nil interface through	1
// TODO: instead of duplicating them, have the compiler decide where symbols go.	1
// TODO: Can we also simplify cases like:	1
MOVD    (R20), R18    // TODO: This turns bad writes into bad reads.	1
// TODO (register args) this is a bit of a hack.	1
// TODO: predecessors is a pain to maintain. Can we somehow order phi	1
// TODO: v = OpPhi (ConstBool [true]) (Arg <bool> {value})	1
for var_ := range s.defvars[b.ID] { // TODO: encode defvars some other way (explicit ops)? make defvars[n] a slice instead of a map.	1
// TODO: Anonymous methods (embedding)	1
// TODO: Benchmark to determine if the pools are necessary. The GC may have	1
// TODO: This loop is O(n^2). It used to be used in nilcheck,	1
/* TODO: MVN C_SHIFT */	1
maxWriteSize uint32 // zero means unlimited; TODO: implement	1
// TODO: Consider removing this check after verifying performance is okay.	1
// TODO: This might be relaxed as in common browsers.	1
// TODO: never rewrite OPANIC to OCALLFUNC in the	1
// TODO: add CloseWrite to crypto/tls.Conn first	1
// TODO: don't use brdist for this case, as it isn't a branch.	1
// TODO: be stricter here? only silently ignore things which	1
// TODO: return true if !v.Type.IsSigned()	1
// TODO: Remove once Makefiles are forgotten.	1
// TODO: Make this be the actual compilation directory, not	1
// TODO: don't dedup pointer and scalar? Rewrite to OpConvert? Can it happen?	1
// TODO does not begin to deal with the full complexity of line numbers.	1
// TODO: Consistency in the encoding would be nice here.	1
// FIXME: the logic of HashStrBytes, HashStrRevBytes, IndexRabinKarpBytes and HashStr, HashStrRev,	1
// TODO: Use index for big endian as found in s390x implementation.	1
// TODO: Are there cases that are safe but don't have a	1
// TODO: Consider passing g.m.p in as an argument so they can be shared	2
// TODO: AT_linkage_name, AT_MIPS_linkage_name.	1
// TODO: BuildN	1
deb.r.Read(b) // TODO: CHECK THESE READS!!	1
// TODO: This could still be a confusing error; maybe goodFunc should provide info.	1
// TODO: wire up necessary VDSO (see os_linux_386.go)	1
// TODO: For big endian implementation, do not reverse bytes.	1
// TODO: generate and print a mapping from nodes to values and blocks	1
// TODO: handle relative paths of recursive expansions in different directories?	1
call = s.newValue0A(ssa.OpStaticLECall, aux.LateExpansionResultType(), aux) // TODO paramResultInfo for NewProc	1
// R4 = TODO: What's this supposed to be?	1
// TODO: what if there's more than one? Probably doesn't matter much.	1
// TODO: once issue 7921 is fixed, this should be reverted to	1
// Calls. TODO not all calls are equal, names give useful clues.	1
// TODO: sort by distance, pick the closest ones?	1
// TODO: optimize. The equivalent code on posix is alloc-free.	1
// TODO: similar mark on non-LR architectures.	1
// TODO: write h12Compare test showing whether	1
// TODO: add support for using the kind	2
// TODO: Auto-generate some day. (Hard-coded in binaries so not likely to change.)	2
roObject, readonly, err := f.Slice(uint64(length)) // TODO: no need to map blocks that are for tools only (e.g. RefName)	1
/* TODO: could be clever, and use general constant builder */	1
// TODO: allow if all indexes are constant.	1
// TODO: Split apart?	1
// TODO: go/doc does not include typed constants in the constants	1
// TODO: use a cast, or is typing directly ok?	1
// TODO: Handle unsigned comparisons?	1
// It is kind complex64 but not type complex64. TODO: We can handle this unsafely.	2
typ.tflag = 0 // TODO: set tflagRegularMemory	1
// TODO: ReleaseTags (need to load default)	1
// TODO: use more compact encoding. The encoding is stateful, and	1
// TODO: get rid of the extra \n inserted by enc.Encode.	1
// TODO inlining produces expressions with ninits. we can't print these yet.	1
// TODO: Consider re-implementing using Advanced SIMD	1
// FIXME: it would be more efficient to bulk-allocate DIEs.	1
// TODO: this is O(N), fix it.	1
// TODO: RFC 3875 isn't clear if only GET is supported, but it	1
// TODO: Col is only called when defining a macro and all it cares about is increasing	1
// TODO: handle more than 2 predecessors, e.g. a || b || c.	2
// TODO: Use a note like we use signals on POSIX OSes	1
// TODO: revert this back to a global map once golang.org/issue/2559 is fixed.	2
// It is kind int16 but not type int16. TODO: We can handle this unsafely.	2
// TODO implement lock striping	1
// TODO: Do a better job yet. Here's one possibility:	1
// TODO: probably no need to export this.	1
// TODO: expand this to all static composite literal nodes?	1
// TODO: handle 32 bit operations	1
q.From.Scale = 2 // TODO: use 1	1
// TODO: expand the set of supported platforms, with testing. Nothing about	1
if aloc := f.RegAlloc[a.ID]; aloc != loc { // TODO: .Equal() instead?	1
// TODO: could be great to move suffix-related things into	1
// TODO: We could pass the entry argument map	1
// TODO: We could narrow this down if the compiler	1
// TODO: move this file to ../ssa, then use sparseSet there.	1
// TODO: Implement smarter auto-zoom using the viewBox attribute	1
// TODO: Remove code looking for the old symbols when we no longer care about 1.3.	2
// TODO: Perhaps we should report no function at all in that case.	1
x := uintptr(pcOff) + datap.text - datap.minpc // TODO: are datap.text and datap.minpc always equal?	1
// TODO: it would really be nicer for internal/cpu to figure out this	1
// TODO: call poll with timeout == 0	1
// TODO: For big endian implementation, the bytes in these slices should be in reverse order,	1
// TODO: Not yet implemented.	1
// FIXME: add support for lowering CondSelects on more architectures	1
// TODO: when the array size is small, unroll the length match checks.	1
name := ldr.RawSymName(s) // TODO: try not to read the name	3
// TODO: double check all gp. shouldn't be getg().	1
// TODO: pprof's remapMappingIDs makes two adjustments:	1
// TODO: maybe post-process output to prevent it from containing	1
# TODO: this assumes tags are just simply comma separated. For now this is all the uses.	2
// TODO returns a non-nil, empty Context. Code should use context.TODO when	1
if atomic.Load(&sched.nmspinning)+atomic.Load(&sched.npidle) == 0 && atomic.Cas(&sched.nmspinning, 0, 1) { // TODO: fast atomic	1
// TODO: We should export a regular C function to panic, change SWIG	1
// TODO: accept N and Ni independently?	1
// TODO: make ITab a separate field in DynamicTypeAssertExpr?	1
// TODO: get rid of OpVarKill here. It's useful for stack frame allocation	1
// TODO: return an error if WriteHeader gets a return parameter	1
if v.Op != ssa.OpStoreReg { // TODO: include other store instructions?	1
MOVD    (R2), R10   // TODO: This turns bad writes into bad reads.	1
// TODO: capture logging during this pass, add it to the HTML	1
Args:     []parse.Node{parse.NewIdentifier(identifier).SetTree(nil).SetPos(pos)}, // TODO: SetTree.	1
// It is kind float32 but not type float32. TODO: We can handle this unsafely.	2
// TODO: Remove this code once we can introduce	1
// TODO: elide nop	1
// TODO fix the nasty storeArgOrLoad recursion in ssa/expand_calls.go so this Just Works with store of a big Arg.	1
//   (See also the TODO below.)	2
// TODO: normalize quotes.	1
// TODO: These should be locals in testAtomic64, but we don't 8-byte	1
// TODO: optionally enforce? Or enforce at the time we receive	1
// TODO: check sym->type against outer->type.	1
// TODO: fix races and enable the following flags	1
// TODO: Check in input, preserve in brchain.	1
// TODO: adjust this writer size to account for frame size +	1
// TODO: are there others?	1
// TODO: check if there are cases that matter where we have	1
saX = sa[x] // TODO bounds check	4
// TODO: This formatting is pretty awful.	1
stackLock    uint32 // sigprof/scang lock; TODO: fold in to atomicstatus	1
if !bytes.Equal(h0, h) { // TODO: constant time?	1
// This implementation only works within a basic block. TODO: use something more global.	1
// FIXME: add check for isExternal?	1
//  Linux         65530 (vm.max_map_count) // TODO: query /proc/sys/vm/max_map_count?	1
// TODO in theory could align offset only to minimum required by spilled data types.	2
MOVW    (R2), R0    // TODO: This turns bad writes into bad reads.	1
// TODO For now, this is only used for debugging output, and if we need more/better information, it might change.	1
// TODO: check for cycles in values	1
// TODO these may be duplicated. Should memoize. Intermediate selectors will go dead, no worries there.	1
// TODO: save marks only for bodies which	1
// It is kind uint16 but not type uint16. TODO: We can handle this unsafely.	2
// TODO: remove file/line from InlinedCall?	1
// TODO: the Microsoft doco says IMAGE_SYM_DTYPE_ARRAY is 3 and IMAGE_SYM_DTYPE_FUNCTION is 2	1
outRegs := noRegisters // TODO if this is costly, hoist and clear incrementally below.	1
// TODO: check if db.driver supports an optional	1
// TODO: figure out whether it's more efficient to just have these	1
// TODO: support multiple text sections.	1
// TODO use this information to make sparse maps faster.	1
// TODO what about arm64 frame pointer adjustment?	1
su.SetRelocAdd(rIdx, int64(braddoff(int32(r.Add()), ldr.SymPlt(targ)/4))) // TODO: don't use r.Add for instruction bytes (issue 19811)	1
// TODO: improve this for HTML?	1
// TODO: if anybody asks, add a Server field, and remember to define the	1
// TODO: This routine should do whatever is needed	1
// TODO: Should scanobject/scanblock just stuff pointers into	1
// TODO: output in sorted order?	1
// TODO: For trivial unsafe.Pointer arithmetic, it would be	1
// TODO: should we special-case cancellation?	1
// TODO: impose length limit	1
// FIXME: this seems clunky, maybe there is a better way to do this.	1
// TODO: do whatever the spec ends up doing.	1
return c.writeTo(b, UDPAddrFromAddrPort(addr)) // TODO: optimize instead of allocating	1
// See also TODO in NeedEmit.	1
// TODO: try to make more variables SSAable?	1
f(REG_V1, 2, REG_V7, 20) // TODO what about V16:31?	1
// TODO: maybe do this for PAUTO as well?	1
// TODO: combine phielim and copyelim into a single pass?	1
// TODO: find a way to make this O(1) without arbitrary cutoffs.	1
// fprintFunc(&buf, f) // TODO: HTML, not text, <br> for line breaks, etc.	1
/*TODO if math.MaxUint32 < x {	2
// FIXME: pass empty string here for name? This would	1
// TODO: keep a single varnum map, then make all of these maps slices instead?	1
// TODO: would it make sense to return an arch-specific	1
// TODO: consider mutating more than one value at a time.	1
B.NE    byte_loop    // unaligned a, use byte-wise compare (TODO: try to align a)	1
// TODO: see writeResHeaders.staysWithinBuffer	1
// TODO: Handle types defined within a function.	1
// showing the TODO body.	1
// TODO if they are not needed for 1.18 / next register abi port, delete them.	1
// TODO: decide whether to a) use this in the client code (which didn't	1
bz2.wantBlockCRC = uint32(br.ReadBits64(32)) // skip checksum. TODO: check it if we can figure out what it is.	1
// TODO: Use maxSize or drop it.	1
// TODO: have the frontend give us branch prediction hints for	1
// TODO: replace pe.NewFile with pe.Load (grep for "add Load function" in debug/pe for details)	1
pendingLines.set(v.Pos, int32(i)) // TODO could be more than one pos for a line	1
// TODO: judging by RFC 7540, Section 6.5.3 each SETTINGS frame should be	1
var errRequestCanceledConn = errors.New("net/http: request canceled while waiting for connection") // TODO: unify?	1
// TODO endianness? Should high-order (sign bits) word come first?	1
// TODO: optimizations. See arm and amd64 LoweredNilCheck.	1
// TODO: export table.	1
// TODO: Enable profiling interrupts.	1
// TODO: Nested simple blocks get unnecessary (but correct) counters	1
// TODO: remove. Use ctxt.Errorf instead.	1
// TODO: Maybe allow contents of a Value to live on the stack.	2
for _, e := range backedges { // TODO: could filter here by calls in loops, if declared and inferred nosplit are recorded in export data.	1
// TODO: Any intra-value formatting?	1
// TODO: update when package unicode captures more of the properties.	2
// TODO: switch to FPXX after be sure that no odd-number-fpr is used.	1
// TODO: regMask -> regSet?	1
// TODO provide a non-nil *Context	2
// TODO: this is a common one. It'd be nice to return true	1
// TODO: move elsewhere?	1
// TODO: Empirically we still need the fd == nil check. Why?	1
// TODO: CGO_FLAGS?	1
// TODO: add support for sharing conns based on cert names	1
// TODO: The second branch is probably predict-not-taken since it is for FP unordered	1
// TODO: For big endian, use index found in s390x implementation.	1
MOVD    R8, R4        // pc being called same TODO as above	1
// TODO: care? unlike http/1, it won't mess up our framing, so it's	1
// TODO: also add this to x/net/http2.Configure Transport, behind	1
// TODO: We could strike a different balance; e.g., saving X0	1
// It is kind int8 but not type int8. TODO: We can handle this unsafely.	2
atomic.Xadd(&sched.npidle, 1) // TODO: fast atomic	1
BNE    byte_loop    // unaligned a, use byte-wise compare (TODO: try to align a)	1
// TODO: A "go:nosplitrec" annotation would be perfect for this.	1
// TODO can this happen recursively?	1
//TODO: remove?  Subsumed by SpillUse?	1
// TODO: the Microsoft doco says IMAGE_SYM_DTYPE_ARRAY is 3 (same with IMAGE_SYM_DTYPE_POINTER and IMAGE_SYM_DTYPE_FUNCTION)	1
// TODO: neeeds to be implemented properly.	2
// TODO: prioritize likely successor?	1
// TODO: avoid the ldr.Lookup calls below by instead using an aux	1
// TODO: try to understand what's really going on.	1
// TODO: could do better allocation-wise here, but trailers are rare,	1
// TODO: Use constraint.IsGoBuild once it exists.	1
//TODO: benchmark and try to find criteria for swapping between	1
// TODO: Use pkgsFilter?	1
// TODO: other attributes?	1
// TODO if roff != 0 { panic }	1
// TODO: Currently our best solution is to find these manually and list them as	1
// TODO: honor GOCLOBBERDEADHASH, or maybe GOSSAHASH.	1
// TODO: other architectures.	1
// TODO: for really big clears, use MOVNTDQ, even without AVX2.	1
// TODO: Don't build a shared library, once SWIG emits the necessary	1
// TODO: send more details to the peer somehow. But http2 has	1
p.Reg = bi // TODO: This is a hack since BI bits are not enumerated as registers	1
// TODO: Anything for tagProfile_KeepFrames?	1
// TODO: validate r.Method too? At least it's less likely to	1
// TODO: Should we elide unexported symbols from a single spec?	1
// TODO: fix test_main to avoid race and improve caching, also allowing to	1
// TODO: keep the result of this function somewhere in the ODOT Node	1
// TODO: find a better place and optimize to only sort TOC symbols	1
// TODO: fix. Maybe mmap the file.	1
// TODO: use CarrierSymByType	1
// TODO: do we need more commands? asm? cgo? For now, no.	1
// TODO: There is no reason we need a generator for this variable, and it	1
// TODO: For big endian, use index values as found in s390x implementation.	1
nx := ir.NewSelectorExpr(base.Pos, ir.OXDOT, np, f.Sym) // TODO: fields from other packages?	1
// TODO: handle errors better.	1
// TODO: If ss doesn't contain any OpPhis, are s0 and s1 dead code anyway.	1
// TODO: pack the first 3 fields more tightly?	1
// TODO: add child task creation events into the parent task events	1
// TODO: suspendG blocks (and spins) until gp	1
// TODO: Pick a better sub-sample of the Plan 9 palette.	1
// TODO: send a RST_STREAM	1
addDFphis(x, h, c, f, defForUses, newphis, sdom) // TODO: convert to explicit stack from recursion.	1
// TODO: add other architectures?	1
} else { // TODO: what's a good status code for canceled requests? 400?	1
// TODO: in efence mode, stack is sysAlloc'd, so this wouldn't	1
// It is kind int32 but not type int32. TODO: We can handle this unsafely.	2
// TODO: This ignores shadowing.	1
BytePtr    *types.Type // TODO: use unsafe.Pointer instead?	1
// TODO: error reporting detail	1
// TODO: fix IMAGE_SYM_DTYPE_ARRAY value and use following expression, instead of 0x0308	1
// TODO -- this needs cleanup; it just works for SSA-able aggregates, and won't fully generalize to register-args aggregates.	1
// TODO: Store filenames as symbols. (Note this would be easiest if you	1
// TODO: pick best of (already processed) predecessors?	1
// TODO: it would be nice if there were a way to run a GC	1
// TODO: FPCR? I don't think we'll change it, so no need to save.	1
// TODO: Issue 26303: close HTTP/2 conns as soon as they become idle.	1
//TODO: maybe remove ReadTuple score above, if it does not help on performance	1
MOVD    R0, 0(R0) // TODO: TD	1
// TODO: do something? External symbols have different representation of attributes.	1
// TODO: maybe do something similar to peimporteddlls to collect all lib names	1
//       // TODO: include ABI version as well?	1
Sym_ *types.Sym // TODO: Remove	1
// FIXME: not clear why it is sometimes necessary to suppress .tbss section creation.	1
// TODO: Using the value ID as the class ignores the fact	2
// TODO if 386 versions are created, also update needSplit and gen/386splitload.rules	1
// TODO: Make unsafe.Pointer the same shape as normal pointers.	1
// TODO not sure if this is best done in wide precision or narrow	1
// TODO: is this really right? Seems to happen a whole lot...	1
// It is kind int64 but not type int64. TODO: We can handle this unsafely.	2
// TODO: this is an artifact of how funcpctab combines information for instructions at a single PC.	1
MOV    (T0), A0    // TODO: This turns bad writes into bad reads.	1
// It is kind uint64 but not type uint64. TODO: We can handle this unsafely.	2
MOVV    (R20), R1    // TODO: This turns bad writes into bad reads.	1
// TODO: This creation of the named function should probably really be done in a	1
result := ir.NewNameAt(pos, typecheck.LookupNum("r", i)) // TODO: names not needed?	1
// TODO: if v is a phi, save desired register for phi inputs.	1
// TODO: we set HasFunctions if all symbols from samples were symbolized (hasFuncs).	1
// TODO: not implemented on ARM	2
// TODO: if the output parameter is heap-allocated, then we	1
// TODO: use MIB_IF_ROW2.AccessType now that we no longer support	1
// TODO: deal with GOAWAY more. particularly the error code	1
// (TODO: derivation?)	1
// TODO: if the req.ContentLength is large, we allocate memory	1
// TODO: how about p->yes->b->yes, i.e. a loop in yes.	1
// TODO: include stack trace of creation time	1
// TODO if roff != 0 { panic }	1
// TODO: delete? replace with something meaningful?	1
// TODO: reuse these slots. They'll need to be erased first.	1
// TODO (aix): make it available	1
// TODO: handle this case? Named return values must be	1
// TODO: central lock to access the map is not ideal.	1
// TODO: expand documentation.	1
// TODO: compact encoding, as above.	1
// TODO: scale the graph with the viewBox attribute.	1
// TODO: We could avoid shading any of the "new" pointers in	1
// TODO REGISTER ABI this needs to be cleaned up.	1
// TODO: Implement auto-completion support.	1
// from being assigned too early. See #14591 and #14762. TODO: allow this.	1
* TODO: hash	1
// TODO: turn LDR of GOT entry into ADR of symbol itself	1
// TODO: It would be nice to combine this with findIndVar.	1
// It is kind %[1]s but not type %[1]s. TODO: We can handle this unsafely.	2
// TODO: use step()	1
// FIXME: avoid having to do name inspections here.	1
// TODO: could also seek minimum position?	1
// TODO: this is only valid for ARMv7+	2
switch ir.CurFunc.Nname.Sym().Name { // TODO: is there a better way than hardcoding the names?	1
// TODO: what if there's more than one? Probably doesn't matter much.	1
s.sp = s.entryNewValue0(ssa.OpSP, types.Types[types.TUINTPTR]) // TODO: use generic pointer type (unsafe.Pointer?) instead	1
// TODO: Modify preference using affinity graph.	1
// TODO: just needs relocation, no need to put in .dynsym	2
// TODO: make sure the linker deduplicates them (see dupok in writeType above).	1
// TODO: Is field guaranteed valid? If so we could avoid this check.	1
// TODO this is not executed because the front-end substitutes a runtime call.	1
// TODO: if the writebarrier pass knows how to reorder stores,	1
// TODO: less copy-happy queue.	1
// TODO: ConnectionError might be overly harsh? Check.	1
// TODO how much of the ABI should be printed?	1
//           See also TODO in go/constant/value.go.	2
// TODO: It would be nice to return an error like "not a module".	1
// This is not a type switch. TODO: Should we use an OSWITCH case here instead of OCASE?	1
// TODO: more type checks?	1
// TODO: this is completely broken. The args passed to newosproc0 (in asm_amd64.s)	1
// TODO: Handle fields that are anonymous structs by	1
// TODO: live at start of block instead?	1
// TODO: fix the order of evaluation, so that the lval of lhs[0]	1
// TODO do them here (or earlier),	1
// TODO what is best arrangement for nested loops?	1
// TODO: Just ignore mem operands?	1
// TODO: handle initial declaration not including an assignment and followed by a single assignment?	1
// TODO: introduce auxSymOff32	1
// FIXME: try converting this map (and the next one) to a single	1
// TODO: debugging - set flags to control verbosity of compiler,	1
// TODO: This may be unnecessary on Windows, which	1
// TODO: are there double-rounding problems here? See issue 48807.	1
# TODO: print interface's methods and dynamic type's func pointers thereof.	1