Russ Cox [Wed, 23 Dec 2020 05:46:27 +0000 (00:46 -0500)]
[dev.regabi] cmd/compile: split out package objw [generated]
Object file writing routines are used not just at the end
of the compilation but also during static data layout in walk.
Split them into their own package.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# Move bit vector to new package bitvec
mv bvec.n bvec.N
mv bvec.b bvec.B
mv bvec BitVec
mv bvalloc New
mv bvbulkalloc NewBulk
mv bulkBvec.next bulkBvec.Next
mv bulkBvec Bulk
mv H0 h0
mv Hp hp
# Leave bvecSet and bitmap hashes behind - not needed as broadly.
mv bvecSet.extractUniqe bvecSet.extractUnique
mv h0 bvecSet bvecSet.grow bvecSet.add \
bvecSet.extractUnique hashbitmap bvset.go
mv bv.go cmd/compile/internal/bitvec
ex . ../arm ../arm64 ../mips ../mips64 ../ppc64 ../s390x ../riscv64 {
import "cmd/internal/obj"
var a *obj.Addr
var i int64
Addrconst(a, i) -> a.SetConst(i)
var p, to *obj.Prog
Patch(p, to) -> p.To.SetTarget(to)
}
rm Addrconst Patch
Russ Cox [Wed, 23 Dec 2020 05:43:42 +0000 (00:43 -0500)]
[dev.regabi] cmd/compile: split up typecheck1 [generated]
typecheck1 is the largest non-machine-generated function in the compiler.
weighing in at 1,747 lines. Since we are destroying the git blame history
anyway, now is a good time to split each different case into its own function,
making future work on this function more manageable.
[git-generate]
cd src/cmd/compile/internal/typecheck
rf '
# Remove tracing print from typecheck1 - the one in typecheck is fine.
# Removing it lets us remove the named result.
# That lets all the cut-out functions not have named results.
rm typecheck.go:/^func typecheck1/+0,/^func typecheck1/+4
sub typecheck.go:/^func typecheck1/+/\(res ir\.Node\)/ ir.Node
Russ Cox [Wed, 23 Dec 2020 05:41:49 +0000 (00:41 -0500)]
[dev.regabi] cmd/compile: split out package typecheck [generated]
This commit splits the typechecking logic into its own package,
the first of a sequence of CLs to break package gc into more
manageable units.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# The binary import/export has to be part of typechecking,
# because we load inlined function bodies lazily, but "exporter"
# should not be. Move that out of bexport.go.
mv exporter exporter.markObject exporter.markType export.go
# Use the typechecking helpers, so that the calls left behind
# in package gc do not need access to ctxExpr etc.
ex {
import "cmd/compile/internal/ir"
# TODO(rsc): Should not be necessary.
avoid TypecheckExpr
avoid TypecheckStmt
avoid TypecheckExprs
avoid TypecheckStmts
avoid TypecheckAssignExpr
avoid TypecheckCallee
Russ Cox [Wed, 23 Dec 2020 05:39:45 +0000 (00:39 -0500)]
[dev.regabi] cmd/compile: move type size calculations into package types [generated]
To break up package gc, we need to put these calculations somewhere
lower in the import graph, either an existing or new package. Package types
already needs this code and is using hacks to get it without an import cycle.
We can remove the hacks and set up for the new package gc by moving the
code into package types itself.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# Remove old import cycle hacks in gc.
rm TypecheckInit:/types.Widthptr =/-0,/types.Dowidth =/+0 \
../ssa/export_test.go:/types.Dowidth =/-+
ex {
import "cmd/compile/internal/types"
types.Widthptr -> Widthptr
types.Dowidth -> dowidth
}
# Disable CalcSize in tests instead of base.Fatalf
sub dowidth:/base.Fatalf\("dowidth without betypeinit"\)/ \
// Assume this is a test. \
return
Russ Cox [Wed, 23 Dec 2020 05:10:25 +0000 (00:10 -0500)]
[dev.regabi] cmd/compile: group known symbols, packages, names [generated]
There are a handful of pre-computed magic symbols known by
package gc, and we need a place to store them.
If we keep them together, the need for type *ir.Name means that
package ir is the lowest package in the import hierarchy that they
can go in. And package ir needs gopkg for methodSymSuffix
(in a later CL), so they can't go any higher either, at least not all together.
So package ir it is.
Rather than dump them all into the top-level package ir
namespace, however, we introduce global structs, Syms, Pkgs, and Names,
and make the known symbols, packages, and names fields of those.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
add go.go:$ \
// Names holds known names. \
var Names struct{} \
\
// Syms holds known symbols. \
var Syms struct {} \
\
// Pkgs holds known packages. \
var Pkgs struct {} \
# Move version check into types.
# A little surprising, but its keyed off the types.Pkg.
ex {
import "cmd/compile/internal/types"
var p *types.Pkg
var major, minor int
langSupported(major, minor, p) -> AllowsGoVersion(p, major, minor)
}
rm langSupported
mv checkLang ParseLangFlag
mv lang langWant AllowsGoVersion ParseLangFlag \
parseLang currentLang goVersionRE goversion.go
# The type1.go here is to avoid an undiagnosed bug in rf
# that does not get the follow-up typechecking right if we
# move directly to type.go during the mv into package types below.
mv \
IsInt IsOrdered IsReflexive \
IsDirectIface IsInterfaceMethod IsMethodApplicable IsPaddedField \
IsRuntimePkg IsReflectPkg ReceiverBaseType \
FloatForComplex ComplexForFloat \
TypeSym TypeSymLookup TypeSymName \
typepkg SimType \
type1.go
# The alg1.go here is because we are only moving part of alg.go.
mv typeHasNoAlg TypeHasNoAlg
mv AlgKind ANOEQ AlgType TypeHasNoAlg IsComparable IncomparableField IsPaddedField alg1.go
Russ Cox [Wed, 23 Dec 2020 05:05:23 +0000 (00:05 -0500)]
[dev.regabi] cmd/compile: move helpers into package base [generated]
[git-generate]
cd src/cmd/compile/internal/gc
rf '
# Move EnableTrace constant into base, with the other flags.
mv enableTrace EnableTrace
mv EnableTrace base.go
Russ Cox [Wed, 23 Dec 2020 05:02:08 +0000 (00:02 -0500)]
[dev.regabi] cmd/compile: remove Left, Right etc methods [generated]
Now that the generic graph structure methods - Left, Right, and so on -
have been removed from the Node interface, each implementation's uses
can be replaced with direct field access, using more specific names,
and the methods themselves can be deleted.
This automated CL adds type assertions on the true branches of
n.Op() equality tests, to redeclare n with a more specific type, when
it is safe to do so. (That is, when n is not reassigned with a more
general type, when n is not reassigned and then used outside the
scope, and so on.) All the "unsafe" times that the automated tool
would avoid have been removed or rewritten in earlier CLs, so that
after this CL and the next one, which removes the use of ir.Nod,
every use of the Left, Right, and so on methods is done using concrete
types, never the Node interface.
Having done that, the CL locks in the progress by deleting many of
the access methods, including Left, SetLeft and so on, from the
Node interface.
There are still uses of Name, Func, Sym, some of the tracking
bits, and a few other miscellaneous fields, but all the main access
methods are gone from the Node interface. The others will be cleaned
up in smaller CLs.
Passes buildall w/ toolstash -cmp.
[git-generate]
cd src/cmd/compile/internal/gc
rf 'typeassert {
import "cmd/compile/internal/ir"
var n ir.Node
Rewrite all uses of ir.Nod and friends to call the IR constructors directly.
This gives the results a more specific type and will play nicely with
introduction of more specific types throughout the code in a followup CL.
Passes buildall w/ toolstash -cmp.
[git-generate]
cd src/cmd/compile/internal/gc
rf '
ex . ../ir {
import "cmd/compile/internal/ir"
import "cmd/compile/internal/types"
import "cmd/compile/internal/syntax"
import "cmd/internal/src"
var p *noder
var orig syntax.Node
var op ir.Op
var l, r ir.Node
var sym *types.Sym
p.nod(orig, op, l, r) -> ir.NodAt(p.pos(orig), op, l, r)
p.nodSym(orig, op, l, sym) -> nodlSym(p.pos(orig), op, l, sym)
var op ir.Op
var l, r ir.Node
ir.Nod(op, l, r) -> ir.NodAt(base.Pos, op, l, r)
var sym *types.Sym
nodSym(op, l, sym) -> nodlSym(base.Pos, op, l, sym)
}
ex . ../ir {
import "cmd/compile/internal/ir"
import "cmd/internal/src"
# rf overlapping match handling is not quite good enough
# for certain nested rewrites, so handle these two - which often contain other ir.NodAt calls - early.
var l, r ir.Node
var xpos src.XPos
ir.NodAt(xpos, ir.OAS, l, r) -> ir.NewAssignStmt(xpos, l, r)
ir.NodAt(xpos, ir.OIF, l, nil) -> ir.NewIfStmt(xpos, l, nil, nil)
}
ex . ../ir {
import "cmd/compile/internal/ir"
import "cmd/compile/internal/types"
import "cmd/internal/src"
var l, r ir.Node
var sym *types.Sym
var xpos src.XPos
We want almost all cmd subdirectories anyway, and relative to the cost
of the rest of toolchain bootstrapping, copying/rewriting a few extra
source files is way cheaper than the engineering cost of forgetting to
maintain these lists as we split out new packages.
While here, also add cmd/internal/archive (and make it compile with Go
1.4) because it'll be needed in subsequent refactorings anyway; and
skip files starting with # (emacs temporary files) and test files
ending with _test.go.
Change-Id: Ic86e680a5fdfaecd617c36d5d04413293b2d6f52
Reviewed-on: https://go-review.googlesource.com/c/go/+/279832
Run-TryBot: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Russ Cox <rsc@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Jason A. Donenfeld [Sun, 8 Nov 2020 10:57:42 +0000 (11:57 +0100)]
cmd/link: handle grouped resource sections
The Go PE linker does not support enough generalized PE logic to
properly handle .rsrc sections gracefully. Instead a few things are
special cased for these. The linker also does not support PE's "grouped
sections" features, in which input objects have several named sections
that are sorted, merged, and renamed in the output file. In the past,
more sophisticated support for resources or for PE features like grouped
sections have not been necessary, as Go's own object formats are pretty
vanilla, and GNU binutils also produces pretty vanilla objects where all
sections are already merged.
However, GNU binutils is lagging with arm support, and here LLVM has
picked up the slack. In particular, LLVM has its own rc/cvtres combo,
which are glued together in mingw LLVM distributions as windres, a
command line compatible tool with binutils' windres, which supports arm
and arm64. But there's a key difference between binutils' windres and
LLVM's windres: the LLVM one uses proper grouped sections.
So, this commit adds grouped sections support for resource sections to
the linker. We don't attempt to plumb generic support for grouped
sections, just as there isn't generic support already for what resources
require. Instead we augment the resource handling logic to deal with
standard two-section resource objects.
We also add a test for this, akin to the current test for more vanilla
binutils resource objects, and make sure that the rsrc tests are always
performed.
Fixes #42866.
Fixes #43182.
Change-Id: I059450021405cdf2ef1c195ddbab3960764ad711
Reviewed-on: https://go-review.googlesource.com/c/go/+/268337
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Alex Brainman <alex.brainman@gmail.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Russ Cox [Tue, 22 Dec 2020 05:07:40 +0000 (00:07 -0500)]
[dev.regabi] cmd/compile: adjust one case in walkexpr
The mid-case n := n.(*ir.AssignExpr) does not lend itself
well to pulling the code into a new function, because n will
be a function argument and will not be redeclarable.
Russ Cox [Mon, 21 Dec 2020 20:10:26 +0000 (15:10 -0500)]
[dev.regabi] cmd/compile: only access Func method on concrete types
Sets up for removing Func from Node interface.
That means that once the Name reorg is done,
which will let us remove Name, Sym, and Val,
Node will be basically a minimal interface.
Cherry Zhang [Mon, 21 Dec 2020 23:41:16 +0000 (18:41 -0500)]
test: trigger SIGSEGV instead of SIGTRAP in issue11656.go
In issue11656.go, it tests that if the runtime can get a
reasonable traceback when it faults at a non-function PC. It does
it by jumping to an address that contains an illegal or trap
instruction. When it traps, the SIGTRAP crashes the runtime.
This CL changes it to use an instruction that triggers SIGSEGV.
This is due to two reasons:
- currently, the handling of bad PC is done by preparePanic,
which is only used for a panicking signal (SIGSEGV, SIGBUS,
SIGFPE), not a fatal signal (e.g. SIGTRAP).
- the test uses defer+recover to get a traceback, which only
works for panicking signals, not fatal signals.
Ideally, we should handle all kinds of faults (SIGSEGV, SIGBUS,
SIGILL, SIGTRAP, etc.) with a nice traceback. I'll leave this
for the future.
This CL also adds RISCV64 support.
Fixes #43283.
Change-Id: I5e0fbf8530cc89d16e05c3257d282bc1d4d03405
Reviewed-on: https://go-review.googlesource.com/c/go/+/279423
Trust: Cherry Zhang <cherryyz@google.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Than McIntosh [Thu, 24 Sep 2020 17:14:46 +0000 (13:14 -0400)]
[dev.regabi] cmd/compile,cmd/link: initial support for ABI wrappers
Add compiler support for emitting ABI wrappers by creating real IR as
opposed to introducing ABI aliases. At the moment these are "no-op"
wrappers in the sense that they make a simple call (using the existing
ABI) to their target. The assumption here is that once late call
expansion can handle both ABI0 and the "new" ABIInternal (register
version), it can expand the call to do the right thing.
Note that the runtime contains functions that do not strictly follow
the rules of the current Go ABI0; this has been handled in most cases
by treating these as ABIInternal instead (these changes have been made
in previous patches).
Generation of ABI wrappers (as opposed to ABI aliases) is currently
gated by GOEXPERIMENT=regabi -- wrapper generation is on by default if
GOEXPERIMENT=regabi is set and off otherwise (but can be turned on
using "-gcflags=all=-abiwrap -ldflags=-abiwrap"). Wrapper generation
currently only workd on AMD64; explicitly enabling wrapper for other
architectures (via the command line) is not supported.
Also in this patch are a few other command line options for debugging
(tracing and/or limiting wrapper creation). These will presumably go
away at some point.
Updates #27539, #40724.
Change-Id: I1ee3226fc15a3c32ca2087b8ef8e41dbe6df4a75
Reviewed-on: https://go-review.googlesource.com/c/go/+/270863
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Than McIntosh <thanm@google.com>
Matthew Dempsky [Fri, 18 Dec 2020 04:17:04 +0000 (20:17 -0800)]
[dev.regabi] cmd/compile: add ir.BasicLit to represent literals
This CL changes so that all literals are represented with a new,
smaller ir.BasicLit type, so that ir.Name is only used to represent
declared constants.
Passes buildall w/ toolstash -cmp.
Change-Id: I4702b8e3fa945617bd05881d7a2be1205f229633
Reviewed-on: https://go-review.googlesource.com/c/go/+/279153
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Russ Cox <rsc@golang.org>
Matthew Dempsky [Fri, 18 Dec 2020 02:47:26 +0000 (18:47 -0800)]
[dev.regabi] cmd/compile: stop reusing Ntype for OSLICELIT length
For OSLICELITs, we were reusing the Ntype field after type checking to
hold the length of the OSLICELIT's backing array. However, Ntype is
only meant for nodes that can represent types. Today, this works only
because we currently use Name for all OLITERAL constants (whether
declared or not), whereas we should be able to represent them more
compactly with a dedicated type that doesn't implement Ntype.
Passes buildall w/ toolstash -cmp.
Change-Id: I385f1d787c41b016f507a5bad9489d59ccfde7f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/279152
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Russ Cox <rsc@golang.org>
Than McIntosh [Wed, 16 Dec 2020 18:45:48 +0000 (13:45 -0500)]
[dev.regabi] runtime/race: adjust test pattern match for ABI wrapper
Adjust the pattern matching in one of the race output test to allow
for the possible introduction of an ABI wrapper. Normally for tests
that match traceback output wrappers are not an issue since they
are screened out by Go's traceback mechanism, but in this case the
race runtime is doing the unwinding, so the wrapper may be visible.
Change-Id: I45413b5c4701d4c28cc760fccc8203493dbe2874
Reviewed-on: https://go-review.googlesource.com/c/go/+/278756
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Than McIntosh <thanm@google.com>
Nikhil Benesch [Wed, 9 Dec 2020 20:14:59 +0000 (15:14 -0500)]
runtime: correct error handling in several FreeBSD syscall wrappers
The FreeBSD syscall convention uses the carry flag to indicate whether
an error has occured. The sys_umtx_op, thr_new, and pipe2 syscall
wrappers were failing to account for this convention and silently
suppressing errors as a result. This commit corrects these wrappers
by copying the pattern used by the other fallible syscall wrappers.
Note that futexsleep1 must now explicitly ignore the ETIMEDOUT error
from sys_umtx_op. Previously ETIMEDOUT was implicitly ignored because
sys_umtx_op never returned an error.
Fixes #43106.
Change-Id: I9c422b87cf4c6d308003bf42c3b419f785578b5d
Reviewed-on: https://go-review.googlesource.com/c/go/+/276892
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Austin Clements <austin@google.com>
Trust: Than McIntosh <thanm@google.com>
Cherry Zhang [Mon, 21 Dec 2020 19:11:02 +0000 (14:11 -0500)]
cmd/pack: treat compiler's -linkobj output as "compiler object"
Treat the compiler's -linkobj output as "compiler object, which
means "pack c" will "see through" the file and add individual
entry to the new archive, instead of the object as a whole.
This is somewhat peculiar. But Go 1.15's cmd/pack does this,
although seemingly accidental. We just do the same. FWIW, it
does make things more consistent with/without -linkobj flag.
Fixes #43271.
Change-Id: I6b2d99256db7ebf0fa430f85afa7464e334f6bcb
Reviewed-on: https://go-review.googlesource.com/c/go/+/279483
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Than McIntosh [Mon, 14 Dec 2020 15:03:37 +0000 (10:03 -0500)]
[dev.regabi] runtime: fix ABI targets in runtime.panic{Index,Slice} shims
Fix up the assembly shim routines runtime.panic{Index,Slice} and
friends so that their tail calls target ABIInternal and not ABI0
functions. This is so as to ensure that these calls don't go through
an ABI0->ABIInternal wrapper (which would throw off the machinery in
the called routines designed to detect whether the violation happened
in the runtime).
Note that when the compiler starts emitting real register calls to
these routines, we'll need to rewrite them to update the arg size and
ensure that args are in the correct registers. For example, the
current shim
TEXT runtime·panicIndex<ABIInternal>(SB),NOSPLIT,$0-16
MOVQ AX, x+0(FP)
MOVQ CX, y+8(FP)
JMP runtime·goPanicIndex<ABIInternal>(SB)
will need to change to
TEXT runtime·panicIndex<ABIInternal>(SB),NOSPLIT,$0
// AX already set up properly
MOVQ CX, BX // second argument expected in BX
JMP runtime·goPanicIndex<ABIInternal>(SB)
Matthew Dempsky [Sat, 19 Dec 2020 04:14:45 +0000 (20:14 -0800)]
[dev.regabi] cmd/compile: stop using ONONAME with Name
This CL changes NewDeclNameAt to take an Op argument to set the Op up
front, and updates all callers to provide the appropriate Op. This
allows dropping the Name.SetOp method.
Rather than creating Names w/ ONONAME earlier and later adding in the
details, this CL changes the import logic to create and add details at
the same time.
Passes buildall w/ toolstash -cmp.
Change-Id: Ifaabade3cef8cd80ddd6644bff79393b934255d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/279313
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 [Sun, 20 Dec 2020 03:26:06 +0000 (19:26 -0800)]
[dev.regabi] cmd/compile: move itabname call out of implements
We only need to call itabname when actually creating the OCONVIFACE
ops, not any time we test whether a type implements an
interface. Additionally, by moving this call out of implements, we
make it purely based on types, which makes it safe to move to package
types.
Does not pass toolstash -cmp, because it shuffles symbol creation
order.
Change-Id: Iea8e0c9374218f4d97b4339020ebd758d051bd03
Reviewed-on: https://go-review.googlesource.com/c/go/+/279333 Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Michael Anthony Knyszek [Mon, 7 Dec 2020 15:11:46 +0000 (15:11 +0000)]
runtime/metrics: add Read examples
This change adds two examples of using the Read function: one that reads
one metric and one that reads all metrics.
Change-Id: I4940a44c9b1d65f3f7a1554e3145ff07e6492fc1
Reviewed-on: https://go-review.googlesource.com/c/go/+/275855
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Michael Pratt <mpratt@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Russ Cox [Mon, 21 Dec 2020 06:44:49 +0000 (01:44 -0500)]
[dev.regabi] cmd/compile: separate ssa from other phases
isIntrinsicCall and ssaDumpInline are the only two "forward references"
to ssa by earlier phases. Make them a bit more explicit so that the
uses and the definitions can end up in different packages.
Change-Id: I02c7a27464fbedef9fee43c0e4094fa08b4d7a5c
Reviewed-on: https://go-review.googlesource.com/c/go/+/279300
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, 21 Dec 2020 06:36:15 +0000 (01:36 -0500)]
[dev.regabi] cmd/compile: separate noder more cleanly
Separate embed, cgo pragmas, and Main trackScopes variable
from noder more cleanly.
This lets us split embed and noder into new packages.
It also assumes that the local embedded variables will be
removed and deletes them now for simplicity.
Change-Id: I9638bcc2c5f0e76440de056c6285b6aa2f73a00d
Reviewed-on: https://go-review.googlesource.com/c/go/+/279299
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, 21 Dec 2020 06:29:02 +0000 (01:29 -0500)]
[dev.regabi] cmd/compile: collect global compilation state
There are various global variables tracking the state of the
compilation. Collect them in a single global struct instead.
The struct definition is in package ir, but the struct itself is
still in package gc. It may eventually be threaded through the
code, but in the short term will end up in package typecheck.
Change-Id: I019db07aaedaed2c9b67dd45a4e138dc6028e54c
Reviewed-on: https://go-review.googlesource.com/c/go/+/279297
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, 21 Dec 2020 06:14:36 +0000 (01:14 -0500)]
[dev.regabi] cmd/compile: setup to move Addrconst, Patch into cmd/internal/obj
Deleting the Pc assignment from Patch is safe because the actual PCs
are not assigned until well after the compiler is done patching jumps.
And it proves that replacing uses of Patch with SetTarget will be safe later.
Change-Id: Iffcbe03f0b5949ccd4c91e79c1272cd06be0f434
Reviewed-on: https://go-review.googlesource.com/c/go/+/279296
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>
inl.go was updated for changes on dev.regabi: namely that OSELRECV has
been removed, and that OSELRECV2 now only uses List, rather than both
Left and List.
order.go was updated IsAutoTmp is now a standalone function, rather
than a method on Node.
ssa.go was similarly updated for new APIs involving package ir.
The tests are all merging upstream additions for gccgo error messages
with changes to cmd/compile's error messages on the dev.regabi branch.
Cherry Zhang [Thu, 12 Nov 2020 02:27:56 +0000 (21:27 -0500)]
runtime: use _exit on darwin
On darwin, where we use libc for syscalls, when the runtime exits,
it calls libc exit function, which may call back into user code,
e.g. invoking functions registered with atexit. In particular, it
may call back into Go. But at this point, the Go runtime is
already exiting, so this wouldn't work.
On non-libc platforms we use exit syscall directly, which doesn't
invoke any callbacks. Use _exit on darwin to achieve the same
behavior.
No test for now, as it doesn't pass on all platforms (see trybot
run of PS2).
May fix #42465.
May fix #43294.
Change-Id: Ia1ada22b5da8cb64fdd598d0541eb90e195367eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/269378
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Michael Pratt [Fri, 11 Dec 2020 19:14:30 +0000 (14:14 -0500)]
runtime: detect netbsd netpoll overrun in sysmon
The netbsd kernel has a bug [1] that occassionally prevents netpoll from
waking with netpollBreak, which could result in missing timers for an
unbounded amount of time, as netpoll can't restart with a shorter delay
when an earlier timer is added.
Prior to CL 232298, sysmon could detect these overrun timers and
manually start an M to run them. With this fallback gone, the bug
actually prevents timer execution indefinitely.
As a workaround, we add back sysmon detection only for netbsd.
Richard Miller [Sat, 5 Dec 2020 19:53:08 +0000 (19:53 +0000)]
runtime: skip wakep call in wakeNetPoller on Plan 9
This was part of a performance improvement made by CL 232298 to
reduce timer latency. On multiprocessor Plan 9 machines, it triggers
memory faults often enough that the builder test suite never completes
successfully. See issue #42303 for discussion. As shown by the benchmark
result below, worst case latency on plan9_arm is very bad even with the
wakep call in place - in the tickers-per-P=1 case, a 3ms timer is 270ms late.
Skipping the wakep call and running the benchmark again shows some cases
worse, some better. The performance cost doesn't seem excessive for this
temporary workaround which makes the plan9_arm builders usable again.
Change-Id: I70c63cb2a2bad46950a7cd9dfc7bb32943710d32
Reviewed-on: https://go-review.googlesource.com/c/go/+/275672 Reviewed-by: David du Colombier <0intro@gmail.com>
Trust: Michael Pratt <mpratt@google.com>
Previously, reassigned was failing to detect reassignments due to
channel receives in select statements (OSELRECV, OSELRECV2), or due to
standalone 2-value receive assignments (OAS2RECV). This was reported
as a devirtualization panic, but could have caused mis-inlining as
well.
Fixes #43292.
Change-Id: Ic8079c20c0587aeacff9596697fdeba80a697b12
Reviewed-on: https://go-review.googlesource.com/c/go/+/279352
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Ian Lance Taylor [Wed, 16 Dec 2020 04:24:33 +0000 (20:24 -0800)]
test: for issue11656 try to execute trap, not call it
The issue11656 code was using the trap instruction as a PC value,
but it is intended to call a PC value that contains the trap instruction.
It doesn't matter too much as in practice the address is not
executable anyhow. But may as well have the code act the way it
is documented to act.
Also, don't run the test with gccgo/GoLLVM, as it can't work.
The illegal instruction will have no unwind data, so the unwinder
won't be able to get past it. In other words, gccgo/GoLLVM suffer
from the exact problem that the issue describes, but it seems insoluble.
For golang/go#11656
Change-Id: Ib2e50ffc91d215fd50e78f742fafe476c92d704e
Reviewed-on: https://go-review.googlesource.com/c/go/+/278473
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Than McIntosh <thanm@google.com> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Ian Lance Taylor [Fri, 18 Dec 2020 21:54:27 +0000 (13:54 -0800)]
test: permit "exponent too large" error
The language spec only requires a signed binary exponent of 16 bits
for floating point constants. Permit a "exponent too large" error for
larger exponents.
Don't run test 11326b with gccgo, as it requires successful compilation
of floating point constants with exponents that don't fit in 16 bits.
Change-Id: I98688160c76864aba525a151a14aaaf86bc36a6f
Reviewed-on: https://go-review.googlesource.com/c/go/+/279252
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Ian Lance Taylor [Thu, 17 Dec 2020 22:59:45 +0000 (14:59 -0800)]
go/build: make TestDependencies work again
CL 243940 accidentally broke TestDependencies such that it always passed.
Make it work again, and add a test so that it won't break in the same way.
This revealed that the new embed package was missing from TestDepencies,
so add it.
Fixes #43249
Change-Id: I02b3e38dd35ad88880c4344d46de13b7639aa4c6
Reviewed-on: https://go-review.googlesource.com/c/go/+/279073
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Ian Lance Taylor [Thu, 17 Dec 2020 22:57:20 +0000 (14:57 -0800)]
os: remove dependency on strings package
Historically the os package has not imported the strings package.
That was enforced by go/build.TestDependencies, but that test
was accidentally broken (#43249). A dependency of os on strings
was accidentally added by CL 266364; remove it.
For #42026
For #43249
Change-Id: If932308f30561fdcc5c608d7563e849c0d2870d8
Reviewed-on: https://go-review.googlesource.com/c/go/+/279072
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
Russ Cox [Thu, 17 Dec 2020 13:49:22 +0000 (08:49 -0500)]
[dev.regabi] cmd/compile: remove prealloc map
The prealloc map seems to exist to avoid adding a field to all nodes.
Now we can add a field to just the nodes that need the field,
so let's do that and avoid having a magic global with extra node state
that isn't preserved by operations like Copy nor printed by Dump.
This also makes clear which nodes can be prealloc'ed.
In particular, the code in walkstmt looked up an entry in
prealloc using an ONAME node, but there's no code that
ever stores such an entry, so the lookup never succeeded.
Having fields makes that kind of thing easier to see and fix.
Passes buildall w/ toolstash -cmp.
Change-Id: I418ad0e2847615c08868120c13ee719dc0b2eacb
Reviewed-on: https://go-review.googlesource.com/c/go/+/278915
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 [Thu, 17 Dec 2020 07:56:26 +0000 (02:56 -0500)]
[dev.regabi] cmd/compile: remove uses of Name.Offset, Name.copy
For globals, Name.Offset is used as a way to address a field within
a global during static initialization. This CL replaces that use with
a separate NameOffsetExpr (ONAMEOFFSET) node.
For locals, Name.Offset is the stack frame offset. This CL calls it
that (FrameOffset, SetFrameOffset).
Now there is no longer any use of Name.Offset or Name.SetOffset.
And now that copies of Names are not being made to change their
offsets, we can lock down use of ir.Copy on Names. The only
remaining uses are during inlining and in handling generic system
functions. At both those times you do want to create a new name
and that can be made explicit by calling the new CloneName method
instead. ir.Copy on a name now panics.
Passes buildall w/ toolstash -cmp.
Change-Id: I0b0a25b9d93aeff7cf4e4025ac53faec7dc8603b
Reviewed-on: https://go-review.googlesource.com/c/go/+/278914
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 [Thu, 10 Dec 2020 23:47:09 +0000 (18:47 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - sinit
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
This CL focuses on sinit.go.
Passes buildall w/ toolstash -cmp.
Change-Id: I3e9458e69a7a9b3f2fe139382bf961bc4473cc42
Reviewed-on: https://go-review.googlesource.com/c/go/+/277928
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 [Thu, 10 Dec 2020 23:45:14 +0000 (18:45 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - walk
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
This CL focuses on walk.go.
Passes buildall w/ toolstash -cmp.
Change-Id: I7aab57e4077cf10da1994625575c5e42ad114a9c
Reviewed-on: https://go-review.googlesource.com/c/go/+/277921
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, 16 Dec 2020 15:53:20 +0000 (10:53 -0500)]
[dev.regabi] cmd/compile: rename ir.Find to ir.Any and update uses
ir.Find is called "any" in C#, Dart, Haskell, Python, R, Ruby, and Rust,
and "any_of" in C++, "anyMatch" in Java, "some" in JavaScript,
"exists in OCaml, and "existsb" in Coq.
(Thanks to Matthew Dempsky for the research.)
This CL changes Find to Any to use the mostly standard name.
It also updates wrapper helpers to use the any terminology:
hasCall -> anyCall
hasCallOrChan -> anyCallOrChan
hasSideEffects -> anySideEffects
Unchanged are "hasNamedResults", "hasUniquePos", and "hasDefaultCase",
which are all about a single node, not any node in the IR tree.
I also renamed hasFall to endsInFallthrough, since its semantics are
neither that of "any" nor that of the remaining "has" functions.
So the new terminology helps separate different kinds of predicates nicely.
Change-Id: I9bb3c9ebf060a30447224be09a5c34ad5244ea0d
Reviewed-on: https://go-review.googlesource.com/c/go/+/278912
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>
Henrique Vicente [Mon, 16 Nov 2020 02:09:31 +0000 (03:09 +0100)]
os/signal: fix flaky tests for NotifyContext.
Test failures started to happen sporadically on some builds after the introduction of NotifyContext.
To make these tests more robust and avoid the risk of crosstalk we run them in a separate process.
Fixes #41561.
Change-Id: Ia7af105c316afd11765358f1e5e253ccfe2adc2b
Reviewed-on: https://go-review.googlesource.com/c/go/+/270198
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Cherry Zhang <cherryyz@google.com>
Than McIntosh [Mon, 14 Dec 2020 18:03:06 +0000 (13:03 -0500)]
cmd/link: avoid use of -T when linking with lld
When doing external linking on Windows, auto-detect the linker flavor
(bfd vs gold vs lld) and when linking with "lld", avoid the use of
"-T" (linker script), since this option is not supported by lld.
[Note: the Go linker currently employs -T to ensure proper placement
of the .debug_gdb_scripts section, to work around issues in older
versions of binutils; LLD recognizes this section and does place it
properly].
Updates #39326.
Change-Id: I3ea79cdceef2316bf86eccdb60188ac3655264ed
Reviewed-on: https://go-review.googlesource.com/c/go/+/278932
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Ian Lance Taylor [Thu, 17 Dec 2020 22:03:07 +0000 (14:03 -0800)]
test: recognize and use gc build tag
Change the run.go driver to recognize the "gc" build tag.
Change existing tests to use the "gc" build tag if they use some
feature that seems specific to the gc compiler, such as passing specific
options to or expecting specific behavior from "go tool compile".
Change tests to use the "!gccgo" build tag if they use "go build" or
"go run", as while those might work with compilers other than gc, they
won't work with the way that gccgo runs its testsuite (which happens
independently of the go command).
For #43252
Change-Id: I666e04b6d7255a77dfc256ee304094e3a6bb15ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/279052
Trust: Ian Lance Taylor <iant@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Dmitri Shuralyov [Wed, 9 Dec 2020 03:54:44 +0000 (22:54 -0500)]
api: promote next to go1.16
Change-Id: Id7d242ddd4b80a763787513d0a658dd7aea9db7d
Reviewed-on: https://go-review.googlesource.com/c/go/+/276454
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Alexander Rakoczy <alex@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org>
Russ Cox [Thu, 10 Dec 2020 23:48:33 +0000 (18:48 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - mop-up
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
This CL handles all the little files that are left.
Russ Cox [Thu, 10 Dec 2020 23:48:18 +0000 (18:48 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - noder
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:47:58 +0000 (18:47 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - subr
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:47:32 +0000 (18:47 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - range, select, swt
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
This CL focuses on range.go, select.go, and swt.go: the big
control structures.
Russ Cox [Thu, 10 Dec 2020 23:46:56 +0000 (18:46 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - const
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:46:45 +0000 (18:46 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - inl
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:46:29 +0000 (18:46 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - import/export
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:46:13 +0000 (18:46 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - escape
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:45:58 +0000 (18:45 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - ssa
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:45:35 +0000 (18:45 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - order
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Thu, 10 Dec 2020 23:42:42 +0000 (18:42 -0500)]
[dev.regabi] cmd/compile: cleanup for concrete types - typecheck
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
Russ Cox [Sat, 12 Dec 2020 23:50:21 +0000 (18:50 -0500)]
[dev.regabi] cmd/compile: simplify ir.Find, replace ir.Inspect with ir.Visit
It seems clear after using these for a week that Find need not return
anything other than a bool saying whether the target was found.
The main reason for not using the boolean earlier was to avoid confusion
with Inspect: for Find, returning true means "it was found! stop walking"
while for Inspect, returning true means "keep walking the children".
But it turns out that none of the uses of Inspect need the boolean.
This makes sense because types can contain expressions, expressions
can contain statements (inside function literals), and so on, so there
are essentially no times when you can say based on the current AST node
that the children are irrelevant to a particular operation.
So this CL makes two changes:
1) Change Find to return a boolean and to take a callback function
returning a boolean. This simplifies all existing calls to Find.
2) Rename Inspect to Visit and change it to take a callback with no
result at all. This simplifies all existing calls to Inspect.
Removing the boolean result from Inspect's callback avoids having
two callbacks with contradictory boolean results in different APIs.
Renaming Inspect to Visit avoids confusion with ast.Inspect.
Russ Cox [Fri, 11 Dec 2020 21:52:21 +0000 (16:52 -0500)]
[dev.regabi] cmd/compile: limit Implicit method to nodes where it is defined
The general concept of an "implicit" operation is provided by every
expr representation, but it really only makes sense for a few of them,
and worse the exact definition of what "implicit" means differs from
node to node.
This CL moves the method to each node implementation, although
they all share the same header bit instead of each defining a bool field
that would turn into 8 bytes on 64-bit systems.
Now we can say precisely which Nodes have a meaningful Implicit
method: AddrExpr, CompLitExpr, ConvExpr, ParenExpr, and StarExpr.
Russ Cox [Fri, 11 Dec 2020 17:55:14 +0000 (12:55 -0500)]
[dev.regabi] cmd/compile: remove use of Initorder, Offset Node fields for initorder
The initorder pass is already making heavy use of maps,
and it is concerned with relatively few nodes (only the assignments
in package-level variable declarations). The tracking of init order
for these nodes can be done with another map instead of storing
the bits directly in the Node representations.
This will let us drop Offset_ from AssignStmt and AssignListStmt
and drop Initorder from all nodes.
Russ Cox [Sat, 12 Dec 2020 02:29:53 +0000 (21:29 -0500)]
[dev.regabi] cmd/compile: change Nodes to be a slice
The Nodes type originally served two purposes:
(1) It provided a representation optimized for empty slices,
allocating only a single word in that case instead of three,
at the cost of a non-empty slice being four words instead of three.
This was particularly important with the old Node representation,
in which most Nodes were full of unused fields.
(2) It provided a few useful helper methods beyond what can be
done with slices.
The downside of Nodes is that the API is a bit overwhelming,
with many ways to spell ordinary slice operations. For example,
reassigning the first node in the list can be done with:
ns.Slice()[0] = n
ns.SetIndex(0, n)
ns.SetFirst(n)
*ns.Addr(0) = n
And APIs must decide whether to use Nodes or []ir.Node and
then conversions must be inserted when crossing the boundary.
Now that Node structs are specialized to opcode and most Nodes
lists are actually non-empty, it makes sense to simplify Nodes
to make it actually a slice type, so that ordinary slice operations can
be used, and assignments can automatically convert between
Nodes and []ir.Node.
This CL changes the representation to be a slice and adds a new
Take method, which returns the old slice and clears the receiver.
In a future CL, the Nodes method set will simplify down to:
Simplifying the API down to just those five methods will also make it
more reasonable to introduce more specialized slices like Exprs and Stmts
at some point in the future.
But again this CL just changes the representation to a slice,
introduces Take, and leaves the rest alone.
Russ Cox [Fri, 11 Dec 2020 01:55:10 +0000 (20:55 -0500)]
[dev.regabi] cmd/compile: update ir/fmt for concrete types
An automated rewrite will add concrete type assertions after
a test of n.Op(), when n can be safely type-asserted
(meaning, n is not reassigned a different type, n is not reassigned
and then used outside the scope of the type assertion,
and so on).
This sequence of CLs handles the code that the automated
rewrite does not: adding specific types to function arguments,
adjusting code not to call n.Left() etc when n may have multiple
representations, and so on.
This CL handles package fmt. There are various type assertions
but also some rewriting to lean more heavily on reflection.
Russ Cox [Mon, 7 Dec 2020 21:07:38 +0000 (16:07 -0500)]
[dev.regabi] cmd/compile: fix potential closure waste in Order
I haven't measured this, but it's the only use of EditChildren
where we aren't careful to allocate a closure once and use it
for the whole recursion. This one is allocating a closure at
every level of the recursion, and it was an oversight that it
wasn't cleaned up in the original CL.
Russ Cox [Mon, 7 Dec 2020 19:58:26 +0000 (14:58 -0500)]
[dev.regabi] cmd/compile: rewrite some generic ir.Nod calls
An automated rewrite is going to remove the bulk of the calls
to ir.Nod and friends. This CL takes care of the ones that don't
have fixed opcodes and so aren't amenable to automatic rewriting.
Passes buildall w/ toolstash -cmp.
Replay of CL 275886, lost to the bad-merge history rewrite.
Russ Cox [Mon, 7 Dec 2020 19:56:49 +0000 (14:56 -0500)]
[dev.regabi] cmd/compile: cleanup preparing for concrete types, 2
Avoid using the same variable for two different concrete
Node types in other files (beyond walk). This will smooth the
introduction of specific constructors, replacing ir.Nod and friends.
Passes buildall w/ toolstash -cmp.
Replay of CL 275885, lost to the bad-merge history rewrite.
Russ Cox [Mon, 7 Dec 2020 19:56:03 +0000 (14:56 -0500)]
[dev.regabi] cmd/compile: cleanup preparing for concrete types
Avoid using the same variable for two different concrete
Node types in walk. This will smooth the introduction of
specific constructors, replacing ir.Nod and friends.
Passes buildall w/ toolstash -cmp.
Replay of CL 275884, lost to the bad-merge history rewrite.
Rob Findley [Wed, 16 Dec 2020 22:19:44 +0000 (17:19 -0500)]
go/types: report error for invalid (but empty) expr switch
This is a port of CL 278132 from the dev.typeparams branch. A notable
addition is a new error code, since no existing codes made sense and we
have an analogous code for type switches.
Fixes #43110
Change-Id: I22b3f9d8777063223f82785504e8b7d299bc5216
Reviewed-on: https://go-review.googlesource.com/c/go/+/278813
Run-TryBot: Robert Findley <rfindley@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Robert Findley <rfindley@google.com>
Cherry Zhang [Mon, 14 Dec 2020 23:52:13 +0000 (18:52 -0500)]
cmd/link: handle large relocation addend on darwin/arm64
Mach-O relocation addend is signed 24-bit. When external linking,
if the addend is larger, we cannot put it directly into a Mach-O
relocation. This CL handles large addend by creating "label"
symbols at sym+0x800000, sym+(0x800000*2), etc., and emitting
Mach-O relocations that target the label symbols with a smaller
addend. The label symbols are generated late (similar to what
we do for RISC-V64).
One complexity comes from handling of carrier symbols, which does
not track its size or its inner symbols. But relocations can
target them. We track them in a side table (similar to what we
do for XCOFF, xcoffUpdateOuterSize).
Fixes #42738.
Change-Id: I8c53ab2397f8b88870d26f00e9026285e5ff5584
Reviewed-on: https://go-review.googlesource.com/c/go/+/278332
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Than McIntosh [Tue, 15 Dec 2020 20:54:25 +0000 (15:54 -0500)]
cmd/internal/goobj: fix buglet in object file reader
The code in the new (introduced in 1.15) Go object file reader was
casting a pointer-mmaped-memory into a large array prior to performing
a read of the relocations section:
For very large object files, this artificial array isn't large enough
(that is, there are more than 1048576 relocs to read), so update the
code to use a larger artifical array size.
Fixes #41621.
Change-Id: Ic047c8aef4f8a3839f2e7e3594bce652ebd6bd5b
Reviewed-on: https://go-review.googlesource.com/c/go/+/278492
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org>
Trust: Than McIntosh <thanm@google.com>