Russ Cox [Wed, 25 Nov 2020 05:30:58 +0000 (00:30 -0500)]
[dev.regabi] cmd/compile: convert types.Node (a pointer) to types.IRNode (an interface)
The pointer hack was nice and saved a word, but it's untenable
in a world where nodes are themselves interfaces with different
underlying types. Bite the bullet and use an interface to hold the
Node when in types.Sym and types.Type.
This has the nice benefit of removing AsTypesNode entirely.
AsNode is still useful because of its nil handling.
Change-Id: I298cba9ff788b956ee287283bec78010e8b601e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/272933
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Russ Cox [Sun, 22 Nov 2020 14:59:15 +0000 (09:59 -0500)]
[dev.regabi] cmd/compile: use Node getters and setters [generated]
Now that we have all the getters and setters defined, use them
and unexport all the actual Node fields. This is the next step
toward replacing Node with an interface.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
ex . ../ir ../ssa {
import "cmd/compile/internal/ir"
import "cmd/compile/internal/types"
import "cmd/internal/src"
var n, x *ir.Node
var op ir.Op
var t *types.Type
var f *ir.Func
var m *ir.Name
var s *types.Sym
var p src.XPos
var i int64
var e uint16
var nodes ir.Nodes
n.Op = op -> n.SetOp(op)
n.Left = x -> n.SetLeft(x)
n.Right = x -> n.SetRight(x)
n.Orig = x -> n.SetOrig(x)
n.Type = t -> n.SetType(t)
n.Func = f -> n.SetFunc(f)
n.Name = m -> n.SetName(m)
n.Sym = s -> n.SetSym(s)
n.Pos = p -> n.SetPos(p)
n.Xoffset = i -> n.SetXoffset(i)
n.Esc = e -> n.SetEsc(e)
# All uses are in other Node methods already.
mv Node.E Node.e
mv Node.Sym Node.sym
mv Node.GetSym Node.Sym
mv Node.Pos Node.pos
mv Node.GetPos Node.Pos
mv Node.Esc Node.esc
mv Node.GetEsc Node.Esc
# While we are here, rename Xoffset to more idiomatic Offset.
mv Node.Xoffset Node.offset
mv Node.GetXoffset Node.Offset
mv Node.SetXoffset Node.SetOffset
# While we are here, rename Ninit, Nbody to more idiomatic Init, Body.
mv Node.Ninit Node.init
mv Node.GetNinit Node.Init
mv Node.PtrNinit Node.PtrInit
mv Node.SetNinit Node.SetInit
mv Node.Nbody Node.body
mv Node.GetNbody Node.Body
mv Node.PtrNbody Node.PtrBody
mv Node.SetNbody Node.SetBody
mv Node.List Node.list
mv Node.GetList Node.List
mv Node.Rlist Node.rlist
mv Node.GetRlist Node.Rlist
# Unexport these
mv Node.SetHasOpt Node.setHasOpt
mv Node.SetHasVal Node.setHasVal
'
The cycle hacks existed because gc needed to import ssa
which need to know about gc.Node. But now that's ir.Node,
and there's no cycle anymore.
Don't know how much it matters but LocalSlot is now
one word shorter than before, because it holds a pointer
instead of an interface for the *Node. That won't last long.
Now that they're not necessary for interface satisfaction,
IsSynthetic and IsAutoTmp can move to top-level ir functions.
Change-Id: Ie511e93466cfa2b17d9a91afc4bd8d53fdb80453
Reviewed-on: https://go-review.googlesource.com/c/go/+/272931
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
If we want to break up package gc at all, we will need to move
the compiler IR it defines into a separate package that can be
imported by packages that gc itself imports. This CL does that.
It also removes the TINT8 etc aliases so that all code is clear
about which package things are coming from.
This CL is automatically generated by the script below.
See the comments in the script for details about the changes.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# These names were never fully qualified
# when the types package was added.
# Do it now, to avoid confusion about where they live.
inline -rm \
Txxx \
TINT8 \
TUINT8 \
TINT16 \
TUINT16 \
TINT32 \
TUINT32 \
TINT64 \
TUINT64 \
TINT \
TUINT \
TUINTPTR \
TCOMPLEX64 \
TCOMPLEX128 \
TFLOAT32 \
TFLOAT64 \
TBOOL \
TPTR \
TFUNC \
TSLICE \
TARRAY \
TSTRUCT \
TCHAN \
TMAP \
TINTER \
TFORW \
TANY \
TSTRING \
TUNSAFEPTR \
TIDEAL \
TNIL \
TBLANK \
TFUNCARGS \
TCHANARGS \
NTYPE \
BADWIDTH
# esc.go and escape.go do not need to be split.
# Append esc.go onto the end of escape.go.
mv esc.go escape.go
# Pull out the type format installation from func Main,
# so it can be carried into package ir.
mv Main:/Sconv.=/-0,/TypeLinkSym/-1 InstallTypeFormats
# Types and fields that need to be exported for use by gc.
mv nowritebarrierrecCallSym SymAndPos
mv SymAndPos.lineno SymAndPos.Pos
mv SymAndPos.target SymAndPos.Sym
# Many methods on Node are actually algorithms that
# would apply to any node implementation.
# Those become plain functions.
mv Node.funcname FuncName
mv Node.isBlank IsBlank
mv Node.isGoConst isGoConst
mv Node.isNil IsNil
mv Node.isParamHeapCopy isParamHeapCopy
mv Node.isParamStackCopy isParamStackCopy
mv Node.isSimpleName isSimpleName
mv Node.mayBeShared MayBeShared
mv Node.pkgFuncName PkgFuncName
mv Node.backingArrayPtrLen backingArrayPtrLen
mv Node.isterminating isTermNode
mv Node.labeledControl labeledControl
mv Nodes.isterminating isTermNodes
mv Nodes.sigerr fmtSignature
mv Node.MethodName methodExprName
mv Node.MethodFunc methodExprFunc
mv Node.IsMethod IsMethod
# Every node will need to implement RawCopy;
# Copy and SepCopy algorithms will use it.
mv Node.rawcopy Node.RawCopy
mv Node.copy Copy
mv Node.sepcopy SepCopy
# Extract Node.Format method body into func FmtNode,
# but leave method wrapper behind.
mv Node.Format:0,$ FmtNode
# Formatting helpers that will apply to all node implementations.
mv Node.Line Line
mv Node.exprfmt exprFmt
mv Node.jconv jconvFmt
mv Node.modeString modeString
mv Node.nconv nconvFmt
mv Node.nodedump nodeDumpFmt
mv Node.nodefmt nodeFmt
mv Node.stmtfmt stmtFmt
# Constant support needed for code moving to ir.
mv okforconst OKForConst
mv vconv FmtConst
mv int64Val Int64Val
mv float64Val Float64Val
mv Node.ValueInterface ConstValue
Russ Cox [Mon, 16 Nov 2020 06:44:47 +0000 (01:44 -0500)]
[dev.regabi] cmd/compile: finish cleanup of Debug parsing
Now that the debug settings are in a struct, use struct tags to set
the usage messages and use reflection to populate debugtab,
much like we did for the Flag struct.
Russ Cox [Mon, 16 Nov 2020 06:17:25 +0000 (01:17 -0500)]
[dev.regabi] cmd/compile: clean up debug flag (-d) handling [generated]
The debug table is not as haphazard as flags, but there are still
a few mismatches between command-line names and variable names.
This CL moves them all into a consistent home (var Debug, like var Flag).
Code updated automatically using the rf command below.
A followup CL will make a few manual cleanups, leaving this CL
completely automated and easier to regenerate during merge
conflicts.
Matthew Dempsky [Wed, 25 Nov 2020 05:56:47 +0000 (21:56 -0800)]
[dev.regabi] cmd/compile: fix reporting of overflow
In the previous CL, I had incorrectly removed one of the error
messages from issue20232.go, because I thought go/constant was just
handling it. But actually the compiler was panicking in nodlit,
because it didn't handle constant.Unknown. So this CL makes it leave
n.Type == nil for unknown constant.Values.
While here, also address #42732 by making sure to report an error
message when origConst is called with an unknown constant.Value (as
can happen when multiplying two floating-point constants overflows).
Finally, add OXOR and OBITNOT to the list of operations to report
errors about, since they're also constant expressions that can produce
a constant with a greater bit length than their operands.
Fixes #42732.
Change-Id: I4a538fbae9b3ac4c553d7de5625dc0c87d9acce3
Reviewed-on: https://go-review.googlesource.com/c/go/+/272928
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Russ Cox [Mon, 16 Nov 2020 05:59:30 +0000 (00:59 -0500)]
[dev.regabi] cmd/compile: clean up flag handling [generated]
The flag values have grown fairly haphazard, with no organization
or even common naming convention. This CL moves all flag values
into the Flag struct (formerly misnamed Debug), except for a few
that live in Ctxt fields instead.
This CL is entirely automated changes.
A followup CL will make a few manual cleanups, leaving this CL
completely automated and easier to regenerate during merge
conflicts.
Cleaning up flags is necessary because the printing routines
look at some of them, and the printing routines need to move
out of package gc to a new package shared by gc and any
other packages that split out of gc.
# Remove //go:generate line copied from main.go
# along with two self-assignments from the merge.
rm flag.go:/go:generate/-+ \
flag.go:/Ctxt.Pkgpath = Ctxt.Pkgpath/-+ \
flag.go:/Ctxt.Debugvlog = Ctxt.Debugvlog/-+
'
Russ Cox [Wed, 25 Nov 2020 04:58:36 +0000 (23:58 -0500)]
[dev.regabi] cmd/compile: simplify fmt handling of Nodes
The existing code introduces many types in what appears to be an
attempt to avoid allocation when converting formatting argument lists.
Simplify by accepting that allocation is going to happen, especially
when Node itself turns into an interface.
Russ Cox [Tue, 17 Nov 2020 16:18:45 +0000 (11:18 -0500)]
[dev.regabi] cmd/compile: add Node field getters and setters
The goal is to move Node to being an interface and then break
up the one big struct into many implementations.
Step 1 is to convert all current uses of Node to only use methods,
so that the existing algorithms keep working even as the underlying
implementations are adjusted.
Step 0 - this CL - is to add the getters and setters for Step 1.
Change-Id: I0570d8727c3ccb64113627bb9bebcb0dc39da07a
Reviewed-on: https://go-review.googlesource.com/c/go/+/273007
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Russ Cox [Sun, 22 Nov 2020 17:09:08 +0000 (12:09 -0500)]
[dev.regabi] cmd/compile: rewrite problematic use of Node fields
For the upcoming rewrite to access methods, a few direct accesses
are problematic for the automated tool, most notably direct copies
or use of Node structs as opposed to pointers.
Russ Cox [Wed, 25 Nov 2020 03:09:57 +0000 (22:09 -0500)]
[dev.regabi] cmd/compile: remove uses of dummy
Per https://developers.google.com/style/inclusive-documentation,
since we are editing some of this code anyway and it is easier
to put the cleanup in a separate CL.
Matthew Dempsky [Tue, 24 Nov 2020 23:52:13 +0000 (15:52 -0800)]
[dev.regabi] add FatalfAt and fix Fatalf docs
I've wanted a FatalfAt function for a while, but under the old "-l"
suffix naming convention it would have been called "Fatalfl", which is
just atrocious.
Change-Id: If87f692ecdff478769426d4b054ac396e5c1e42e
Reviewed-on: https://go-review.googlesource.com/c/go/+/273013
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Matthew Dempsky [Sat, 14 Nov 2020 07:36:48 +0000 (23:36 -0800)]
[dev.regabi] cmd/compile: replace Val with go/constant.Value
This replaces the compiler's legacy constant representation with
go/constant, which is used by go/types. This should ease integrating
with the new go/types-based type checker in the future.
Performance difference is mixed, but there's still room for
improvement.
Russ Cox [Tue, 24 Nov 2020 14:37:54 +0000 (09:37 -0500)]
[dev.regabi] cmd/compile: replace evconst with non-mutating version
evconst is one of the largest sources of Op rewrites,
which prevent separating different kinds of nodes
(in this case, arithmetic nodes and OLITERAL nodes).
The change in swt.go is necessary because otherwise
the syntax graph ends up containing that OLEN expression
multiple times, which violates the invariant that it's a tree
except for ONAME, OLITERAL, and OTYPE nodes.
(Before, the OLEN was overwritten by an OLITERAL, so the
invariant still held, but now that we don't overwrite it,
we need a different copy for each instance.)
Passes toolstash -cmp.
Change-Id: Ia004774ab6852fb384805d0f9f9f234b40842811
Reviewed-on: https://go-review.googlesource.com/c/go/+/272869
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Russ Cox [Wed, 18 Nov 2020 16:25:29 +0000 (11:25 -0500)]
[dev.regabi] cmd/compile: introduce OMETHEXPR instead of overloading ONAME
A method expression today is an ONAME that has none of the
invariants or properties of other ONAMEs and is always a special case
(hence the Node.IsMethodExpression method).
Remove the special cases by making a separate Op.
Passes toolstash -cmp.
Change-Id: I7667693c9155d5486a6924dbf75ebb59891c4afc
Reviewed-on: https://go-review.googlesource.com/c/go/+/272867
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Russ Cox [Tue, 24 Nov 2020 15:25:41 +0000 (10:25 -0500)]
[dev.regabi] cmd/compile: add OMETHEXPR
This CL is obviously OK but does not pass toolstash -cmp,
because it renumbers the Op codes. In a separate CL so that
we can use toolstash -cmp on the CL with real changes
related to OMETHEXPR.
Change-Id: I1db978e3f2652b3bdf51f7981a3ba5137641c8c7
Reviewed-on: https://go-review.googlesource.com/c/go/+/272866
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Russ Cox [Mon, 16 Nov 2020 22:00:10 +0000 (17:00 -0500)]
[dev.regabi] cmd/compile: clean up Node.Func
The original meaning of type Func was "extra fields factored out
of a few cases of type Node having to do with functions",
but those specific cases didn't necessarily have any relation.
A typical declared function is represented by an ODCLFUNC Node
at its declaration and an ONAME node at its uses, and both those
have a .Func field, but they are *different* Funcs.
Similarly, a closure is represented both by an OCLOSURE Node for
the value itself and an ODCLFUNC Node for the underlying function
implementing the closure. Those too have *different* Funcs,
and the Func.Closure field in one points to the other and vice versa.
This has led to no end of confusion over the years.
This CL elevates type Func to be the canonical identifier for
a given Go function.
This looks like a trivial CL but in fact is the result of a lot of
scaffolding and rewriting, discarded once the result was achieved, to
separate out the three different kinds of Func nodes into three
separate fields, limited in use to each specific Node type, to
understand which Func fields are used by which Node types and what the
possible overlaps are. There were a few overlaps, most notably around
closures, which led to more fields being added to type Func to keep
them separate even though there is now a single Func instead of two
different ones for each function.
A future CL can and should change Curfn to be a *Func instead of
a *Node, finally eliminating the confusion about whether Curfn
is an ODCLFUNC node (as it is most of the time) or an ONAME node
(as it is when type-checking an inlined function body).
Although sizeof_test.go makes it look like Func is growing by two
words, there are now half as many Funcs in a running compilation,
so the memory footprint has actually been reduced substantially.
Change-Id: I598bd96c95728093dc769a835d48f2154a406a61
Reviewed-on: https://go-review.googlesource.com/c/go/+/272253
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Mon, 23 Nov 2020 21:42:43 +0000 (13:42 -0800)]
[dev.regabi] cmd/compile: prep refactoring for switching to go/constant
This CL replaces gc.Ctype (along with its CTINT, etc. constants) with
constant.Kind; renames Val.Ctype to Val.Kind; and replaces a handful
of abstraction-violating patterns that can be readily expressed
differently.
The next commit will actually replace Val with constant.Value.
Change-Id: I6c22ec0597508845f88eee639a0d76cbaa66d08f
Reviewed-on: https://go-review.googlesource.com/c/go/+/272653
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Tue, 24 Nov 2020 05:48:38 +0000 (21:48 -0800)]
[dev.regabi] strconv: add to bootstrap packages
go/constant relies on strconv for parsing Go literals, while older
versions of strconv either lack recent Go language features (e.g., Go
1.13's new numeric literals) or have errors (e.g., mishandling of
carriage returns in raw string literals prior to Go 1.8).
This requires two changes:
1. Splitting out the internal/bytealg dependency into a separate file,
which can be easily substituted with a simple loop for bootstrap
builds.
2. Updating eisel_lemire.go to not utilize Go 1.13 functionality
(underscores in numeric literals and signed shift counts).
Change-Id: Ib48a858a03b155eebdcd08d577aec2254337e70e
Reviewed-on: https://go-review.googlesource.com/c/go/+/272749 Reviewed-by: Robert Griesemer <gri@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Sat, 14 Nov 2020 02:33:19 +0000 (18:33 -0800)]
[dev.regabi] cmd/compile: remove CTRUNE
Since CL 255217, we've been able to rely on types.UntypedRune to
identify untyped rune literals, rather than needing Mpint.Rune /
CTRUNE. This makes way for switching to using go/constant, which
doesn't have a separate notion of rune constants distinct from integer
constants.
Passes toolstash-check.
Change-Id: I319861f4758aeea17345c101b167cb307e706a0e
Reviewed-on: https://go-review.googlesource.com/c/go/+/272652 Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Sat, 14 Nov 2020 04:38:21 +0000 (20:38 -0800)]
[dev.regabi] cmd/compile: replace CTNIL with ONIL
Properly speaking, "nil" is a zero value, not a constant. So
go/constant does not have a representation for it. To allow replacing
Val with constant.Value, we split out ONIL separately from OLITERAL so
we can get rid of CTNIL.
Passes toolstash-check.
Change-Id: I4c8e60cae3b3c91bbac43b3b0cf2a4ade028d6cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/272650
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Subsequent CL will make use of ONIL. Split out separately so that the
next CL can pass toolstash-check.
Change-Id: I49d77bedbe2cac4a5da149c925cda969e50b0b2d
Reviewed-on: https://go-review.googlesource.com/c/go/+/272649
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Mon, 16 Nov 2020 16:44:40 +0000 (08:44 -0800)]
[dev.regabi] cmd/compile: cleanup type switch typechecking
Address outstanding TODO, which simplifies subsequent CLs.
Now the compiler always type checks type-switch case clauses (like
gccgo), but it treats clause variables as broken if an appropriate
type cannot be determined for it (like go/types).
Passes toolstash-check.
Change-Id: Iedfe9cdf38c6865211e4b93391f1cf72c1bed136
Reviewed-on: https://go-review.googlesource.com/c/go/+/272648
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Mon, 23 Nov 2020 12:28:25 +0000 (04:28 -0800)]
[dev.regabi] go/constant: avoid heap allocations in match
When type switching from interface{} to T, and then returning the T as
interface{} again, it's better to return the original interface{}
value. This avoids needing to heap allocate the T for
non-pointer-shaped types (i.e., int64Val, complexVal, stringVal).
Change-Id: I25c83b3f9ec9bd2ffeec5a65279b68f4fcef8a19
Reviewed-on: https://go-review.googlesource.com/c/go/+/272647
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Tue, 17 Nov 2020 20:53:34 +0000 (12:53 -0800)]
[dev.regabi] go/constant: optimize BitLen
Avoids an unnecessary heap allocation when computing the bit length of
int64 values.
Change-Id: I69dfc510e461daf3e83b0b7b6c0707f6526a32d0
Reviewed-on: https://go-review.googlesource.com/c/go/+/272646
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
Russ Cox [Mon, 16 Nov 2020 17:18:09 +0000 (12:18 -0500)]
[dev.regabi] cmd/compile: prepare for package ir
The next CL will introduce a package ir to hold the IR definitions.
This CL adjusts a few names and makes a few other minor changes
to make the next CL - an automated one - smoother.
Russ Cox [Mon, 16 Nov 2020 16:08:38 +0000 (11:08 -0500)]
[dev.regabi] cmd/compile: clean up error API
Prepare for factoring the error API out of this package by
cleaning it up. The doc comments use the intended new names,
which will be introduced in the next CL.
Russ Cox [Sun, 15 Nov 2020 16:40:25 +0000 (11:40 -0500)]
[dev.regabi] cmd/compile: rewrite concurrentFlagOk to be clearer
The current implementation copies Debug, clears a bunch of flags
that are meant to be considered OK, and then checks the result
against the zero value. But more flags are cleared than remain:
it's easier to write and to understand to just check the ones that
need checking.
This phrasing also makes it safe to move more flags into the struct.
It turns out that some of the flags being checked should probably
not be checked, but this CL is meant to be a strict semantic no-op,
so left a TODO to clean up the function a bit more later.
Now that there's no code remaining that uses Func.Nname, we can get
rid of it along with the remaining code that uselessly assigns to it.
Passes toolstash-check.
Change-Id: I104ab3bb5122fb824c741bc6e4d9d54fefe5646e
Reviewed-on: https://go-review.googlesource.com/c/go/+/272390
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Mon, 23 Nov 2020 04:43:16 +0000 (20:43 -0800)]
[dev.regabi] cmd/compile/internal/gc: refactor to use stop using Func.Nname
Automated factoring produced by rf script below to replace uses of
Func.Nname with Field.Nname or Node.MethodName as appropriate.
Some dead assignments to Func.Nname are left behind; these will be
removed in a subequent remove-only CL.
Passes toolstash-check.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
ex \
import "cmd/compile/internal/types"; \
var f *types.Field; \
var n *types.Node; \
f.Type.Nname() -> f.Nname; \
f.Type.SetNname(n) -> f.Nname = n; \
f.Type.FuncType().Nname -> f.Nname
ex \
var n *Node; \
asNode(n.Type.Nname()) -> n.MethodName(); \
asNode(n.Type.FuncType().Nname) -> n.MethodName(); \
asNode(callpartMethod(n).Type.Nname()) -> n.MethodName()
'
Change-Id: Iaae054324dfe7da6f5d8b8d57a1e05b58cc5968c
Reviewed-on: https://go-review.googlesource.com/c/go/+/272389
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Sun, 22 Nov 2020 18:45:44 +0000 (10:45 -0800)]
[dev.regabi] cmd/compile/internal/gc: prep for Func.Nname removal refactoring
There are three bits of method-handling code where we separately go
from Field->Type and then Type->Node. By shuffling the code around a
little to go Field->Type->Node in a single statement, we're able to
more easily remove Type from the operation.
Passes toolstash-check.
Change-Id: Ife98216d70d3b867fa153449abef0e56a4fb242a
Reviewed-on: https://go-review.googlesource.com/c/go/+/272388
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Sun, 22 Nov 2020 21:47:55 +0000 (13:47 -0800)]
[dev.regabi] cmd/compile/internal/gc: add MethodName for getting referenced method
A common operation throughout the front end is getting the ONAME for a
method used in a method selector, method expression, or method value.
This CL adds MethodName as a uniform API for doing this for all of
these kinds of nodes.
For method selectors (ODOTMETH) and method expressions (ONAMEs where
isMethodExpression reports true), we take advantage of the Node.Opt
field to save the types.Field. This is the approach we already started
taking in golang.org/cl/271217 (caching types.Field in Node.Opt for
ODOT).
For method values (OCALLPART), we continue using the existing
callpartMethod helper function. Escape analysis already uses Node.Opt
for tracking the method value's closure's data flow.
A subsequent, automated refactoring CL will make more use of this
method. For now, we just address a few cases in inl.go that aren't
easily automated.
Passes toolstash-check.
Change-Id: Ic92b288b2d8b2fa7e18e3b68634326b8ef0d869b
Reviewed-on: https://go-review.googlesource.com/c/go/+/272387
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Mon, 23 Nov 2020 08:15:40 +0000 (00:15 -0800)]
[dev.regabi] cmd/compile/internal/types: add pos/sym/typ params to NewField
These are almost always set, so might as well expect callers to
provide them. They're also all required by go/types's corresponding
New{Field,Func,Param,Var} functions, so this eases API compatibility.
Passes toolstash-check.
Change-Id: Ib3fa355d4961243cd285b41915e87652ae2c22f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/272386
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com> Reviewed-by: Robert Griesemer <gri@golang.org>
The bug was introduced by https://golang.org/cl/220844.
Fixes #42790.
Change-Id: I44d619a1a4d3f2aee1c5575d5cfddcc4ba10895f
Reviewed-on: https://go-review.googlesource.com/c/go/+/272666
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Michael Anthony Knyszek [Mon, 23 Nov 2020 19:25:32 +0000 (19:25 +0000)]
runtime/metrics: tweak wording of stack and unused memory metrics
This change tweaks and simplifies the descriptions of a couple metrics
to make them easier to parse (for humans).
Change-Id: I852654c7e7042c662ebdfa6334e3baf49ca4b33c
Reviewed-on: https://go-review.googlesource.com/c/go/+/272566
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com> Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Martin Möhrmann [Sat, 21 Nov 2020 14:53:18 +0000 (15:53 +0100)]
sync: use 386 instead of x86-32 to refer to the 32 bit x86 architecture
This aligns the naming with GOARCH using 386 as a build target for
this architecture and makes it more easily found when searching
for documentation related to the build target.
Change-Id: I393bb89dd2f71e568124107b13e1b288fbd0c76a
Reviewed-on: https://go-review.googlesource.com/c/go/+/271988
Trust: Martin Möhrmann <moehrmann@google.com>
Run-TryBot: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Keith Randall [Fri, 20 Nov 2020 21:59:29 +0000 (13:59 -0800)]
cmd/compile: add test for 42753
This issue was already fixed at tip. Just adding the test that
failed on 1.14/1.15.
Update #42753
Change-Id: I00d13ade476b9c17190d762d7fdcb30cf6c83954
Reviewed-on: https://go-review.googlesource.com/c/go/+/272029
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Jason A. Donenfeld [Fri, 13 Nov 2020 14:48:05 +0000 (15:48 +0100)]
syscall: add DLLError.Unwrap function
Because we're expecting for future functions to be unavailable, we
should add an Unwrap() function to the DLLError struct, so that people
can test for this situation easily via:
if errors.Is(err, syscall.ERROR_PROC_NOT_FOUND) { ... }
DLLError already was wrapping the underlying Errno error, but never got
the Go 1.13 helper method.
Fixes golang/go#42584
Change-Id: I0f32a5146946b1b37a30897ba825a56faefc792c
Reviewed-on: https://go-review.googlesource.com/c/go/+/269761
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Trust: Alex Brainman <alex.brainman@gmail.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Bryan C. Mills [Fri, 20 Nov 2020 22:03:42 +0000 (17:03 -0500)]
cmd/go/internal/modload: remove the Reqs function
The Reqs function returns an mvs.Reqs implemention for the global
build list. The API that it presents assumes that the build list is
globally consistent (problematic for #40775) and readily available
(problematic for #36460).
Fortunately, it is no longer used outside of the modload package.
We can instead use individual instances of the unexported mvsReqs
struct, making the dependency on the global build list more explicit.
For #36460
For #40775
Change-Id: I8674442f2a86416b0bf9c3395cb591c1e724c9d2
Reviewed-on: https://go-review.googlesource.com/c/go/+/272129
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Michael Matloob [Fri, 20 Nov 2020 19:09:03 +0000 (14:09 -0500)]
cmd/go: support the -overlay flag for go mod commands
Move the declaration of the -overlay flag to base.AddModCommonFlags,
where other flags that are needed for go mod commands and for builds
are declared. The flag's already initialized in modload.Init so
there's no additional work needed to be done to support it in the go
mod commands.
For #39958
Change-Id: I70725d620cc69cb820f6ed923d626f4fe041b1c5
Reviewed-on: https://go-review.googlesource.com/c/go/+/272126
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Jay Conrod <jayconrod@google.com>
Ian Lance Taylor [Fri, 20 Nov 2020 20:54:18 +0000 (12:54 -0800)]
cmd/cgo, cmd/internal/pkgpath: support gofrontend mangler v3
The gofrontend mangling scheme used by gccgo and GoLLVM has changed again.
Support the new version. This is a port of the relevant parts of
https://golang.org/cl/271726.
For #41862
Change-Id: I9c961c8e17ec960a83a23e1d49ea900962b63393
Reviewed-on: https://go-review.googlesource.com/c/go/+/272127
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Michael Matloob [Fri, 30 Oct 2020 21:11:36 +0000 (17:11 -0400)]
cmd/go: support overlaying go.mod files
This change updates the lockedfile package to open files using the
new fsys.OpenFile function. The logic of fsys.Open has been moved into
fsys.OpenFile, and fsys.Open is now just a light wrapper around it.
For #39958
Change-Id: I552f1a45ac00ac06b5812008d17a61e610b4b113
Reviewed-on: https://go-review.googlesource.com/c/go/+/266797
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Jay Conrod [Thu, 19 Nov 2020 15:56:35 +0000 (10:56 -0500)]
cmd/go: remove ListModules call in runGet
ListModules was used to download .info files so that 'go list -m all'
would succeed later when offline. However, 'go list -m all' may
already fail when offline after 'go mod tidy', so it doesn't make
sense to add complexity to 'go get'.
Instead, remove the ListModules call and fix the test that
accidentally depended on it.
For #42723
Change-Id: I692597cf5ca15c23fa6fc9d2bac4b6e044299482
Reviewed-on: https://go-review.googlesource.com/c/go/+/271577
Trust: Jay Conrod <jayconrod@google.com> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Bryan C. Mills [Thu, 19 Nov 2020 18:58:52 +0000 (13:58 -0500)]
cmd/go/internal/modload: eliminate LoadedModules
As of CL 271646, all external callers have been eliminated. Replace
the remaining internal caller with a direct reference to the buildList
variable and remove the exported function to prevent backsliding.
For #36460
Change-Id: Iea82df1e3e604ada602dda3e830c06d441eee2a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/271647
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Jay Conrod <jayconrod@google.com>
Bryan C. Mills [Thu, 19 Nov 2020 18:26:51 +0000 (13:26 -0500)]
cmd/go/internal/work: remove a redundant call to modload.LoadedModules
The modload.EditBuildList call added in CL 270980 already ensures that
installMod does not require a newer version of itself, so the condition
that this loop is checking for is redundant.
(I had meant for this change to be included in CL 270980, but
apparently somehow reverted it prior to mailing.)
For #36460
Change-Id: I4dd746b927f7012d950187cac9c510cd6fec8fd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/271646
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Trust: Bryan C. Mills <bcmills@google.com>
Ian Lance Taylor [Fri, 20 Nov 2020 03:15:57 +0000 (19:15 -0800)]
net, internal/poll: reset value before adding in minor kernel version
Fixes #42733
Change-Id: I5446aeb5de13cd70212755fb12c9bc484f343c74
Reviewed-on: https://go-review.googlesource.com/c/go/+/271846
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Robert Griesemer [Fri, 20 Nov 2020 02:00:13 +0000 (18:00 -0800)]
go/types: use correct error position
Follow-up on https://golang.org/cl/271706 .
Change-Id: I90339987aed88b0de3ee7ebe7d413282055c260c
Reviewed-on: https://go-review.googlesource.com/c/go/+/271789
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Robert Griesemer [Fri, 20 Nov 2020 00:38:11 +0000 (16:38 -0800)]
go/types: fix error message for consistency
Follow-up on https://golang.org/cl/271706 .
(Missed a review comment.)
Change-Id: Ibff542f43d721600a2452907c0a20941961e793f
Reviewed-on: https://go-review.googlesource.com/c/go/+/271766
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Robert Griesemer [Thu, 19 Nov 2020 20:40:19 +0000 (12:40 -0800)]
go/types, go/constant: handle infinities as unknown values
With this change, constant literals (and results of constant
operations) that internally become infinities are represented
externally (to go/constant) as "unknown" values.
The language has no provisions to deal with infinite constants,
and producing unknown values allows the typechecker to report
errors and avoid invalid operations (such as multiplication of
zero with infinity).
Fixes #20583.
Change-Id: I12f36a17d262ff7957b0d3880241b5a8b2984777
Reviewed-on: https://go-review.googlesource.com/c/go/+/271706
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Findley <rfindley@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Michael Anthony Knyszek [Thu, 19 Nov 2020 15:07:06 +0000 (15:07 +0000)]
runtime/metrics: clarify Read's documentation
Change-Id: Idbcbc304f1568399a82af9dcd51e511393ed5ee0
Reviewed-on: https://go-review.googlesource.com/c/go/+/271558
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com> Reviewed-by: Michael Pratt <mpratt@google.com>
Michael Anthony Knyszek [Thu, 19 Nov 2020 15:01:34 +0000 (15:01 +0000)]
runtime/metrics: clarify memory and GC metrics documentation
Change-Id: I8940990a591a808ddd4b8613531f52453f85bde1
Reviewed-on: https://go-review.googlesource.com/c/go/+/271557
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com> Reviewed-by: Michael Pratt <mpratt@google.com>
Bryan C. Mills [Wed, 18 Nov 2020 18:44:39 +0000 (13:44 -0500)]
cmd/go/internal/work: avoid modload.Selected in 'go install pkg@version'
At this point in installOutsideModule the build list is empty, so
Selected trivially returns "none" for all modules.
(This change could have been made in CL 266657, but it was a bit
simpler to update the QueryPattern call sites mechanically to ensure
that there would be no unintentional semantic drift.)
For #36460
Change-Id: I44fb73794985bfeebb1dde0c092313f319c2945a
Reviewed-on: https://go-review.googlesource.com/c/go/+/271419
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Bryan C. Mills [Wed, 18 Nov 2020 03:54:13 +0000 (22:54 -0500)]
cmd/go: fix failing gccgo cases in TestScript/build_overlay
The 'go install' command does not support the -gccgo flag.
(I'm not sure why, but it doesn't.)
gccgo also uses system-native assembly syntax instead of cmd/compile's
Plan 9 derivative. I've added an assembly file that seems to work on
Linux, but I haven't tested it on other platforms; if it fails on
other platforms, we can refine the test as needed.
Fixes #42688
Change-Id: I0693a6a9eb58975f20cdc4160ef5f9a948563c88
Reviewed-on: https://go-review.googlesource.com/c/go/+/270978
Trust: Bryan C. Mills <bcmills@google.com> Reviewed-by: Michael Matloob <matloob@golang.org>
Dan Scales [Thu, 19 Nov 2020 00:42:31 +0000 (16:42 -0800)]
cmd/compile: flag "-d=dumpptrs" to print Node ptrs in Dump output
The printing of the ptr values can mean that two dump outputs can't easily be
compared for the identical structure, so adding the "-d=dumpptrs" option to make
printing of Node pointer values be an option.
Change-Id: I0e92b02f069e9de2e6fa036a7841645d13cdd7a9
Reviewed-on: https://go-review.googlesource.com/c/go/+/271339
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Jason A. Donenfeld [Wed, 18 Nov 2020 20:38:57 +0000 (21:38 +0100)]
runtime: support new callbackasm1 calling convention on windows/arm
This updates the callbacks implementation on windows/arm for the
changes made in CL 258938. At the time, that was left as a TODO.
At the same time, it also extends the previous support for only 4
arguments to also support additional arguments on the stack. This is
required for functions like SetWinEventHook, which take 7 arguments. It
does this by pushing r0-r3 onto the stack before the normal prologue,
and then pointing the args struct to that location.
This is derived from CL 270077 and CL 270078.
Updates #40724.
Fixes #42591.
Change-Id: Icc199e7f2c24205e41be4e00015283c7e2a9b797
Reviewed-on: https://go-review.googlesource.com/c/go/+/271178
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com> Reviewed-by: Austin Clements <austin@google.com>
Bryan C. Mills [Wed, 18 Nov 2020 04:16:30 +0000 (23:16 -0500)]
cmd/go/internal/modload: remove SetBuildList
For the last remaining call site (in cmd/go/internal/work, added for
the new 'go install pkg@version' codepath in CL 254365), use
EditBuildList instead.
SetBuildList assumes that the caller has enough information to produce
a complete, coherent build list. With lazy loading, producing a
complete, coherent build list is no longer quite so trivial.
In CL 263267, I rewrote the main caller of SetBuildList (the 'go get'
command), and in the process added a more targeted modload hook
(EditBuildList). That hook also suffices for 'go install pkg@version'.
The resulting error messages are perhaps not as smooth as they ought
to be, but if they are too awkward we should probably fix them for
'go get' too, and the commands can continue to share the edit hook.
For #36460
Updates #40276
Change-Id: I698a9dcd2efe6378a4d91f21362880aa8e50001b
Reviewed-on: https://go-review.googlesource.com/c/go/+/270980
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Bryan C. Mills [Wed, 18 Nov 2020 03:13:35 +0000 (22:13 -0500)]
cmd/go/internal/modcmd: eliminate a call to modload.LoadedModules
modload.LoadedModules reveals more information than necessary about
whether modules have been loaded lazily. The 'vendor' subcommand
doesn't actually need that much information: it has all of the
information that it needs from prior calls to LoadPackages and
ModFile.
For #36460
Change-Id: If08733cca930b2b80616b037b63985ecfd6a320b
Reviewed-on: https://go-review.googlesource.com/c/go/+/270979
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Hanlin Shi [Mon, 16 Nov 2020 17:29:33 +0000 (17:29 +0000)]
internal/fmtsort: sort the unsafe pointers in map
Currently storing keys that contain unsafe.
Pointer in a map could result inruntime panic when printing the map.
The root cause is that unsafe.Pointer is not comparable.
Fixes #42622.
Change-Id: Ie3bae7ee4945041843b66514de6227212a3da73e
GitHub-Last-Rev: d12d41302e6118cb457aafb05f7aaed9df259b56
GitHub-Pull-Request: golang/go#42623
Reviewed-on: https://go-review.googlesource.com/c/go/+/270277
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Bryan C. Mills <bcmills@google.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Robert Griesemer [Thu, 19 Nov 2020 01:05:02 +0000 (17:05 -0800)]
go/types: report an error for invalid constant values
The parser reports syntactic errors in constant literals.
The go/constant package produces an "unknown" value for
syntactically correct numeric constants that are too small
or too large. Check for the unknown value and report an
error rather than silently continuing.
Fixes #42695.
Change-Id: I414214559a285d67ed50184dc750f106960b5620
Reviewed-on: https://go-review.googlesource.com/c/go/+/271377
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Matthew Dempsky [Wed, 18 Nov 2020 20:50:46 +0000 (12:50 -0800)]
cmd/compile: fix miscompilation during inlining
When inlining a function call expression, it's possible that the
function callee subexpression has side effects that need to be
preserved. This used to not be an issue, because inlining wouldn't
recognize these as inlinable anyway. But golang.org/cl/266199 extended
the inlining logic to recognize more cases, but did not notice that
the actual inlining code was discarding side effects.
Issue identified by danscales@.
Fixes #42703.
Change-Id: I95f8fc076b6ca4e9362e80ec26dad9d87a5bc44a
Reviewed-on: https://go-review.googlesource.com/c/go/+/271219 Reviewed-by: Dan Scales <danscales@google.com>
Trust: Dan Scales <danscales@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Wed, 18 Nov 2020 20:08:59 +0000 (12:08 -0800)]
cmd/compile: fix panic in field tracking logic
Within the frontend, we generally don't guarantee uniqueness of
anonymous types. For example, each struct type literal gets
represented by its own types.Type instance.
However, the field tracking code was using the struct type as a map
key. This broke in golang.org/cl/256457, because that CL started
changing the inlined parameter variables from using the types.Type of
the declared parameter to that of the call site argument. These are
always identical types (e.g., types.Identical would report true), but
they can be different pointer values, causing the map lookup to fail.
The easiest fix is to simply get rid of the map and instead use
Node.Opt for tracking the types.Field. To mitigate against more latent
field tracking failures (e.g., if any other code were to start trying
to use Opt on ODOT/ODOTPTR fields), we store this field
unconditionally. I also expect having the types.Field will be useful
to other frontend code in the future.
Finally, to make it easier to test field tracking without having to
run make.bash with GOEXPERIMENT=fieldtrack, this commit adds a
-d=fieldtrack flag as an alternative way to enable field tracking
within the compiler. See also #42681.
Fixes #42686.
Change-Id: I6923d206d5e2cab1e6798cba36cae96c1eeaea55
Reviewed-on: https://go-review.googlesource.com/c/go/+/271217
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Michael Pratt [Tue, 17 Nov 2020 21:47:08 +0000 (16:47 -0500)]
sync: document RWMutex race semantics
RWMutex provides explicit acquire/release synchronization events to the
race detector to model the mutex. It disables sync events within the
methods to avoid e.g., the atomics from adding false synchronization
events, which could cause false negatives in the race detector.
Change-Id: I5126ce2efaab151811ac264864aab1fa025a4aaf
Reviewed-on: https://go-review.googlesource.com/c/go/+/270865
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com>
Trust: Michael Pratt <mpratt@google.com>
Michael Pratt [Mon, 9 Nov 2020 21:37:05 +0000 (16:37 -0500)]
runtime: give test child time to block
The child in TestPanicSystemstack prints "x\n" and then blocks on a
lock. Receiving those bytes only indicates that the child is _about to
block_. Since we don't have a way to know when it is fully blocked,
sleep a bit to give it time to block. This makes us less likely to lose
the race and signal before the child blocks, which will fail the test as
the stack trace cannot be read from a running G.
Fixes #33626
Change-Id: I8a27b1b114bf75e1e5bcb2a7a33aa69cdbc22f40
Reviewed-on: https://go-review.googlesource.com/c/go/+/268578
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Tobias Klauser [Wed, 18 Nov 2020 10:08:43 +0000 (11:08 +0100)]
runtime: use clock_gettime instead of gettimeofday on darwin
clock_gettime has higher resolution than gettimeofday and is available
since macOS 10.12. Go 1.15 already requires at least macOS 10.12 and
thus clock_gettime can be used unconditionally (also see
https://golang.org/doc/go1.15#darwin)
Fixes #25633
Change-Id: I46305387212735e5d3a13e5f02ec90f3e6d546a4
Reviewed-on: https://go-review.googlesource.com/c/go/+/270918
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Tobias Klauser [Wed, 18 Nov 2020 07:53:00 +0000 (08:53 +0100)]
runtime: use pipe2 syscall for Pipe in tests
On FreeBSD >= 11 with a kernel built with COMPAT_FREEBSD11 but not
COMPAT_FREEBSD10, the pipe syscall is not available. Thus, tests using
runtime.pipe fail with ENOSYS. As suggested by Ian, fix this by calling
pipe2(0) in these tests and fall back to pipe() in case of ENOSYS.
Fixes #42659
Change-Id: Ifbb8008884b7901fe87830d162ad326122c5fab9
Reviewed-on: https://go-review.googlesource.com/c/go/+/270917
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Jay Conrod [Tue, 10 Nov 2020 20:48:37 +0000 (15:48 -0500)]
cmd/go: in 'go get', only load retractions for resolved versions
Previously, 'go get' loaded retractions for every module in the build
list, which took a long time and usually wasn't helpful.
This rolls forward CL 269019, which was reverted in CL 270521. The new
revision adds a call to modload.ListModules at the end of 'go get' to
ensure .info files are cached for everything in the build list.
Fixes #42185
Change-Id: I684f66c5e674384d5a0176fbc8317e5530b8a915
Reviewed-on: https://go-review.googlesource.com/c/go/+/270858
Trust: Jay Conrod <jayconrod@google.com> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Alberto Donizetti [Wed, 28 Oct 2020 09:48:17 +0000 (10:48 +0100)]
time: in NewTicker, document that the 1st tick comes after d
Fixes #42245
Change-Id: I3b298ab6be65569389873d68bd3c6e49cf892c69
Reviewed-on: https://go-review.googlesource.com/c/go/+/265818 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Rob Pike <r@golang.org>
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Robert Griesemer [Tue, 17 Nov 2020 06:23:48 +0000 (22:23 -0800)]
go/types: add test case for incorrect map index expression
The existing code for map index expressions checked the
wrong variable (x rather than key) to see if the index
assignment was correct. Since x.mode was always valid in
that case, type-checking didn't follow the error exit in
case of an incorrect map index expression.
However, since we know the correct map element type
irrespective of the validity of the map key, the existing
code path is preferrable over exiting early via an error
because the map index expression returns a valid type which
then can be used for further type-checking.
Removed the unneeded 'if' statement and added a test case
producing the expected two errors (rather than only one if
we would "correct" the 'if' statement instead).
In summary, this commit adds a test but doesn't change the
behavior of type-checking of map index expressions.
Change-Id: I67845bfaa03600c9400f9a1462d7a68a66921ad4
Reviewed-on: https://go-review.googlesource.com/c/go/+/270658
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
Than McIntosh [Mon, 17 Aug 2020 18:17:07 +0000 (14:17 -0400)]
cmd/compile: clean up buggy DWARF inlined info PC ranges
Repair the code that generates PC ranges for DWARF inlined routine
instances to insure that if II Y is a child of II X within the inline
tree, X's ranges include the ranges from Y. This is similar to what
we're already doing for DWARF scopes.
Updates #33188.
Change-Id: I9bb552777fcd1ae93dc01872707667ad092b1dd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/248724
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> Reviewed-by: David Chase <drchase@google.com>
Trust: Than McIntosh <thanm@google.com>
Dmitri Shuralyov [Tue, 17 Nov 2020 02:04:34 +0000 (02:04 +0000)]
cmd/go: revert "in 'go get', only load retractions for resolved versions"
This reverts CL 269019.
Reason for revert: The TestScript/mod_gonoproxy test is failing
on linux-386-longtest and linux-amd64-longtest builders.
Change-Id: I7e132fb4fb5a9c00add28e5100a0e96a9250282c
Reviewed-on: https://go-review.googlesource.com/c/go/+/270521
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Jay Conrod <jayconrod@google.com>
Lynn Boger [Mon, 16 Nov 2020 14:40:45 +0000 (09:40 -0500)]
cmd/compile: fix rules regression with shifts on PPC64
Some rules for PPC64 were checking for a case
where a shift followed by an 'and' of a mask could
be lowered, depending on the format of the mask. The
function to verify if the mask was valid for this purpose
was not checking if the mask was 0 which we don't want to
allow. This case can happen if previous optimizations
resulted in that mask value.
This fixes isPPC64ValidShiftMask to check for a mask of 0 and return
false.
This also adds a codegen testcase to verify it doesn't try to
match the rules in the future.
Fixes #42610
Change-Id: I565d94e88495f51321ab365d6388c01e791b4dbb
Reviewed-on: https://go-review.googlesource.com/c/go/+/270358
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Paul Murphy <murp@ibm.com> Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Jay Conrod [Mon, 12 Oct 2020 16:03:43 +0000 (12:03 -0400)]
cmd/go: update 'go help mod init'
'go help mod init' now mentions that the module path can be derived
from the directory within GOPATH. We no longer mention version
control, since that's now ignored.
Fixes #36775
Change-Id: Ia5559ecb537fccd838eeab84517e76aa01989292
Reviewed-on: https://go-review.googlesource.com/c/go/+/261539
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Jay Conrod <jayconrod@google.com> Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Michael Matloob <matloob@golang.org>
Joel Sing [Wed, 11 Nov 2020 17:30:15 +0000 (04:30 +1100)]
cmd/link/internal/ld: dedup shared libraries on openbsd
When linking internally on OpenBSD, dedup libraries treating versioned
and unversioned libraries as equivalents. Versioned libraries are preferred
and are retained over unversioned libraries.
This avoids the situation where the use of cgo results in a DT_NEEDED for a
versioned library (for example, libc.so.96.1), while a dynamic import
specifies an unversioned library (for example, libc.so). Without deduplication
this would result in two DT_NEEDED entries, causing a failure when ld.so
attempts to load the Go binrary.
Updates #36435
Fixes #39257
Change-Id: I4a4942f259dece01d97bb51df9e13d67c9f94d34
Reviewed-on: https://go-review.googlesource.com/c/go/+/249978
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>