Robert Findley [Wed, 10 Nov 2021 21:58:45 +0000 (16:58 -0500)]
go/types: rename types.context to types.environment
Now that we have a Context type the context (unexported) type is
particularly confusing. Rename it to environment.
Change-Id: I7d280439b8263d9ebfd561fc4d59c6d43c8d3e3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/363176
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 21:52:17 +0000 (16:52 -0500)]
go/types: don't set a Config.Context if none is provided
Users can re-use a type checking context by passing it via types.Config.
There is no need for us to expose the internal type checking context
when the config context is unset, and in fact doing so could lead to a
memory leak for users that re-use types.Config, expecting it to be small
and immutable.
Keep track of the Context on Checker instead, and zero it out at the end
of type checking.
Change-Id: Iff5b328a09cd0af76fcd4869f5f15352131b5986
Reviewed-on: https://go-review.googlesource.com/c/go/+/363175
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 20:54:00 +0000 (15:54 -0500)]
go/types: when type hashing, canonicalize interfaces
The interface type string preserves certain non-semantic attributes of
the type, such as embedded interfaces. We want the hash to represent the
interface identity, so hash the type set representation of the interface
instead.
Change-Id: I14081ac20b738c5fe11785e0846a9b4358594768
Reviewed-on: https://go-review.googlesource.com/c/go/+/363115
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 20:53:16 +0000 (15:53 -0500)]
go/types: when type hashing, use placeholders for type parameters
Type parameter names don't matter for the purposes of generic type
identity, so mask them with numeric placeholders when hashing.
Change-Id: Iacb4c23abecdd733fc292ae13ecac6baa2c5524c
Reviewed-on: https://go-review.googlesource.com/c/go/+/363114
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 03:31:59 +0000 (22:31 -0500)]
go/types: deduplicate signatures with the context
Extend the type checking context to allow de-duplicating *Signature
instances, in addition to *Named instances.
Naively we would deduplicate instances of different-but-identical origin
*Signature types. That may be OK, but it seems a bit strange to get the
same signature when instantiating two different functions. For now,
differentiate *Signature types by prepending a unique identifier for the
origin pointer, thus guaranteeing that instances de-duplicated if they
come from the exact same (pointer identical) origin type.
Updates #47103
Change-Id: I93cc3cacad195267fe0a5801f9c5a3b1e61eb907
Reviewed-on: https://go-review.googlesource.com/c/go/+/362801
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Griesemer [Fri, 12 Nov 2021 16:59:49 +0000 (08:59 -0800)]
cmd/compile/internal/types2: remove tparamIsIface flag and corresponding dead code
Added/clarified some comments.
Change-Id: Ib08d3343ff08c23cc8880a27a0148d1ff077a80f
Reviewed-on: https://go-review.googlesource.com/c/go/+/363654
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>
Robert Griesemer [Fri, 12 Nov 2021 06:21:48 +0000 (22:21 -0800)]
cmd/compile/internal/types2: remove asNamed
In the few remaining places where we use asNamed, if the argument
is indeed a *Named, we either don't need to look "inside" it, or
we call under() (which calls Named.underlying() which does resolve);
so there's no need for an implicit resolution (which was done by
asNamed). The only place where we do need to resolve is in lookup,
so added the explicit resolve call in that case.
Change-Id: Iff0a19fde7581e94149e89b9e48157c1981db105
Reviewed-on: https://go-review.googlesource.com/c/go/+/363441
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>
Robert Griesemer [Wed, 10 Nov 2021 23:50:35 +0000 (15:50 -0800)]
cmd/compile/internal/types2: set tparamsIsIface to true
This CL enables the mode in which the underlying type of
type parameters is the underlying type of their constraints.
Change-Id: Id3471578dab098695dbd1e0429356ebcc9c5e224
Reviewed-on: https://go-review.googlesource.com/c/go/+/363155
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>
Robert Griesemer [Tue, 26 Oct 2021 17:06:55 +0000 (10:06 -0700)]
cmd/compile/internal/types2: underlying type of a type parameter is its constraint interface
Until now, the type checker operated with the definition that the
underlying type of a type parameter is itself. This leads to some
inconcistencies and caused us to disallow type declarations where
the RHS is a stand-alone type parameter.
This change implements an alernative definition: the underlying
type of a type parameter is the underlying type of its constraint;
i.e., the underlying type of a type parameter is always an interface
(because constraints must be interfaces). This matches the theory
closely and also resolves some inconsistencies. For example, we
don't need to prohibit stand-alone type parameters on the RHS of
a type declaration (though, for the sake of keeping the tests the
same, we still do in this CL). We also get a clear understanding of
what it would mean to use a type assertion or type switch on a type
parameter (still disabled with this CL). Finally, the declaration
of a type parameter now very closely matches the definition of an
ordinary type.
The main consequence is that the rules for assignment need to be
slightly modified: even though a type parameter is an interface,
we cannot simply assign to it per the rules for interfaces: the
type parameter's type is fixed for the instantiation and we need
to reflect that accordingly when checking for assignability.
This CL does not enable the new mode, it implements it in parallel
to the existing mode; the internal flag tparamIsIface is used to
switch between the modes.
The changes to the code are numerous, but straight-forward: when-
ever we deal with an underlying type that might be a type parameter
(or newly, an interface), we need to act slightly differently. For
the time being this leads to some code duplication because the code
supports both modes.
While some of the code for the new mode seems more complicated
(e.g., when we have an interface, the code checks that it is not
the underlying type of a type parameter), in reality many of the
extra checks are redundant and only present because of an abundance
of caution: interfaces with specific type sets are not permitted as
types for ordinary variables, and so even if we were to hit those
cases w/o excluding type parameters the behavior would be the same.
Runs all tests with tparamIsIface enabled and disabled.
Current setting: disabled.
Change-Id: I7bb6453f4fe2569d92face222058fb4e17b12f25
Reviewed-on: https://go-review.googlesource.com/c/go/+/359016
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>
Robert Griesemer [Fri, 12 Nov 2021 21:23:45 +0000 (13:23 -0800)]
cmd/compile/internal/types2: make sure we are safe for nil in underIs
Reviewed all uses of underIs (global function and method) and made
sure we are ok with a nil incoming argument (indicating a type set
with no specific types).
Added a couple of checks where we didn't have them (and somehow
didn't run into a problem yet).
Change-Id: Ifde45a3a80ddf2b1a19c83f79258ad8207dfb09f
Reviewed-on: https://go-review.googlesource.com/c/go/+/363658
Trust: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
Dan Scales [Thu, 11 Nov 2021 18:28:17 +0000 (10:28 -0800)]
cmd/compile: add missing method info for method with correct name except for case
When being used by the compiler, augment the types2 missing method
message with extra info, if a method is missing, but a method with the
correct name except for case (i.e. equal via string.EqualFold()) is
present. In that case, print out the wanted method and the method that
is present (that has the wrong case).
In the 1.17 compiler, we don't do this case-folding check when assigning
an interface to an interface, so I didn't add that check, but we could
add that.
Fixes #48471
Change-Id: Ic54549c1f66297c9221d979d49c1daa719aa66cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/363437
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Dan Scales [Wed, 10 Nov 2021 16:41:21 +0000 (08:41 -0800)]
cmd/compile: match Go 1.17 compiler error messages more closely
When being used by the compiler, fix up types2 error messages to be more
like Go 1.17 compiler errors. In particular:
- add information about which method is missing when a type is not
assignable/convertible/etc. to an interface.
- add information about any existing method which has the same name,
but wrong type.
- add extra hint in the case that the source or destination type is a
pointer to an interface, rather than an interface.
- add extra hint "need type assertion" in the case that the source is
an interface that is implemented by the destination.
- the following change in the CL stack also adds information about any
existing method with a different name that only differs in case.
Include much of the new logic in a new common function
(*Checker).missingMethodReason().
types2 still adds a little more information in some cases then the Go
1.17 compiler. For example, it typically says "(value of type T)",
rather than "(type T)", where "value" could also be "constant",
"variable", etc.
I kept the types2 error messages almost all the same when types2 is not
used by the compiler. The only change (to reduce amount of compatibility
code) was to change "M method" phrasing in one case to "method M"
phrasing in one error message (which is the phrasing it uses in all
other cases). That is the reason that there are a few small changes in
types2/testdata/check/*.src.
Added new test test/fixedbugs/issue48471.go to test that the added
information is appearing correctly.
Also adjusted the pattern matching in a bunch of other
test/fixedbugs/*.go, now that types2 is producing error messages closer
to Go 1.17. Was able to remove a couple test files from the types2
exception list in run.go.
Updated #48471
Change-Id: I8af1eae6eb8a5541d8ea20b66f494e2e795e1956
Reviewed-on: https://go-review.googlesource.com/c/go/+/363436
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Griesemer [Fri, 12 Nov 2021 00:02:35 +0000 (16:02 -0800)]
cmd/compile/internal/types2: remove asTypeParam and simplify some code
Because we do not permit a stand-alone type parameter on the RHS of
a type declaration, the underlying type of a (Named) type cannot be
a type parameter. This allows us to simplify some code.
Specifically, when parsing union elements, we don't need to delay
a check for later, which allows further simplifications when computing
type sets.
Change-Id: I4047c609f87ebb194ea8c1bad630a70d255b20cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/363438
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>
Michael Pratt [Wed, 10 Nov 2021 17:56:40 +0000 (12:56 -0500)]
runtime: start ARM atomic kernel helper traceback in caller
Like the VDSO, we cannot directly traceback from the Linux kernel ARM
atomic/barrier helpers. However, unlike the VDSO, this functions are
extremely simple. Neither of the functions we use, kuser_cmpxchg and
kuser_memory_barrier, touch SP or LR.
We can use this to our advantage to read LR and simply start tracebacks
in the caller.
Fixes #49182
Change-Id: I890edbeb7c128938000fe7baf6f913c02a956edd
Reviewed-on: https://go-review.googlesource.com/c/go/+/362977
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Mui <cherryyz@google.com>
Damien Neil [Mon, 8 Nov 2021 19:23:27 +0000 (11:23 -0800)]
net/http: do not cancel request context on response body read
When sending a Request with a non-context deadline, we create a
context with a timeout. This context is canceled when closing the
response body, and also if a read from the response body returns
an error (including io.EOF).
Cancelling the context in Response.Body.Read interferes with the
HTTP/2 client cleaning up after a request is completed, and is
unnecessary: The user should always close the body, the impact
from not canceling the context is minor (the context timer leaks
until it fires).
Fixes #49366.
Change-Id: Ieaed866116916261d9079f71d8fea7a7b303b8fb
Reviewed-on: https://go-review.googlesource.com/c/go/+/361919
Trust: Damien Neil <dneil@google.com>
Run-TryBot: Damien Neil <dneil@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Michael Pratt [Wed, 10 Nov 2021 22:35:13 +0000 (17:35 -0500)]
runtime/pprof: mark TestCPUProfileMultithreadMagnitude as flaky
The Linux kernel starting in 5.9 and fixed in 5.16 has a bug that can
break CPU timer signal delivery on new new threads if the timer
interrupt fires during handling of the clone system call.
Broken CPU timer signal deliver will skew CPU profile results and cause
this test to fail.
There is currently no known workaround, so mark the test as flaky on
builders with known broken kernels.
For #49065
Change-Id: I37ceb9ea244869b0aab5cd9a36b27ca2f7e5d315
Reviewed-on: https://go-review.googlesource.com/c/go/+/363214
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Michael Pratt [Fri, 12 Nov 2021 16:16:43 +0000 (11:16 -0500)]
runtime: drop cgoTraceback call assumptions from CgoPprof tests
the CgoPprof tests currently assume that calls to their cgoTraceback
functions are primarily for generating pprof samples and exit early
after receiving two calls.
This is a fragile assumption, as cgoTraceback will be called for _any_
signal received, hence why the test already looks for 2 calls instead of
1.
Still, this has caused flaky failures in two cases:
* #37201, where async preemption signals add additional probability of
receiving non-profiling signals. This was resolved by disabling async
preemption.
* #49401, where some ITIMER_PROF SIGPROF signals are ignored in favor of
per-thread SIGPROF signals.
Rather than attempting to keep plugging holes, this CL drops the fragile
assumption from these tests. Now they simply unconditionally run for the
full 1s before exiting.
Fixes #49401
Change-Id: I16dc9d2f16c2fb511e9db93dd096a402121f86ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/363634
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Rhys Hiltner <rhys@justin.tv>
Katie Hockman [Wed, 10 Nov 2021 21:22:08 +0000 (16:22 -0500)]
internal/fuzz: set timeout for each exec of fuzz target
This change sets a timeout of 10 seconds on each
execution of the fuzz target, both during fuzzing
and during minimization. This is not currently
customizable by the user, but issue #48157 tracks
this work.
Deadlocks will be considered non-recoverable errors,
and as such, will not be minimizable.
Context.TypeHash is not being used outside of the type checker, so
unexport it.
The TypeHash method is meant to hash instances, not arbitrary types, and
will soon be modified to differentiate origin types by pointer identity
(even if they are *Signature types).
Change-Id: Ia8d4a7c6350ce7f278b70630585efb0009fef63a
Reviewed-on: https://go-review.googlesource.com/c/go/+/363516
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 03:28:18 +0000 (22:28 -0500)]
go/types: re-use type hashing logic in Context.typeHash
The special handling for *Named types is not necessary. The hash of an
instance is simply the hash of its type followed by its type argument
list.
Change-Id: I7aa58e73b81731c3cad3a2fd14124f63cfb685a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/362800
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 03:20:18 +0000 (22:20 -0500)]
go/types: refactor the Context type map to accept arbitrary types
In preparation for storing *Signature types in Context, refactor the
type map to not depend on the *Named type API.
Change-Id: I0439d43aa4cc3a60a78f409a773a343a4fffd0fa
Reviewed-on: https://go-review.googlesource.com/c/go/+/362799
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Wed, 10 Nov 2021 02:39:53 +0000 (21:39 -0500)]
go/types: use Identical to verify type identity in the Context map
We don't have guarantees that our type hash is perfect, and in fact
fuzzing found cases where identical types hashed to different values. In
case non-identical types hash to the same value, we should ensure that
we de-duplicate using Identical.
Adjust the type map to keep a slice of distinct type identities, so that
we can guarantee that type identity is preserved by de-duplication.
To allow look-up of instances by their identity, before they are
actually instantiated, add a Context.lookup method that accepts origin
type and type arguments. Replace the multi-function typeForHash method
with an update method that requires its argument be non-nil.
Change-Id: I8fe6fb2955f508db608161b7285b02d0a2fa0e46
Reviewed-on: https://go-review.googlesource.com/c/go/+/362798
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Michael Anthony Knyszek [Fri, 12 Nov 2021 05:04:32 +0000 (05:04 +0000)]
test/recover4.go: use mprotect to create a hole instead of munmap
Currently the recover4 test, which recovers from a panic created from a
fault, generates a fault by creating a hole in a mapping. It does this
via munmap. However, it's possible the runtime can create a new mapping
that ends up in that hole, for example if the GC executes, causing the
test to fail.
In fact, this is the case now with a smaller minimum heap size.
Modify the test to use mprotect, and clean up the code a little while
we're here: define everything in terms of the length of original
mapping, deduplicate some constants and expressions, and have the test
recover properly even if recover() returns nil (right now it panics
because it fails to type assert nil as error).
Fixes #49381.
Change-Id: If399eca564466e5e8aeb2dc6f86a246d0fce7b5d
Reviewed-on: https://go-review.googlesource.com/c/go/+/363534
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Mui <cherryyz@google.com>
Cherry Mui [Fri, 12 Nov 2021 00:58:23 +0000 (19:58 -0500)]
reflect: keep pointer in aggregate-typed args live in Call
When register ABI is used, reflect.Value.Call prepares the call
arguments in a memory representation of the argument registers.
It has special handling to keep the pointers in arguments live.
Currently, this handles pointer-typed arguments. But when an
argument is an aggregate-type that contains pointers and passed
in registers, it currently doesn't keep the pointers live. Do
so in this CL.
May fix #49363.
Change-Id: Ic6a0c5fdf9375ef02f7c03fbe9345e2e98c9353d
Reviewed-on: https://go-review.googlesource.com/c/go/+/363358
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Cherry Mui [Thu, 11 Nov 2021 21:51:08 +0000 (16:51 -0500)]
cmd/link: don't unmap output file at error exit
When the link exits on error it currently calls Out.Close, which
will munmap the output buffer and close the file. This may be
called in concurrent phase where other goroutines may be writing
to the output buffer. The munmap can race with the write, causing
it to write to unmapped memory and crash. This CL changes it to
just close the file without unmapping. We're exiting on error
anyway so no need to unmap.
Fixes #47816.
Change-Id: I0e89aca991bdada3d017b7d5c8efc29e46308c03
Reviewed-on: https://go-review.googlesource.com/c/go/+/363357
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Pedro Lopez Mareque [Fri, 8 Oct 2021 05:01:35 +0000 (07:01 +0200)]
math/bits: add examples for Add, Sub, Mul and Div
Change-Id: I2f3619aa827e18f356871511c20cf2c712f496b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/353189
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Keith Randall <khr@golang.org>
Katie Hockman [Tue, 9 Nov 2021 22:13:36 +0000 (17:13 -0500)]
all: update terminology for fuzzing
This change doesn't modify any functionality.
It also doesn't update all of the comments and
variable names of the internal code, but everything
user facing should be correct.
Updates #49185
Change-Id: Ia8b2c94b89ba45897c4085ea0c17a3d8896f7ec7
Reviewed-on: https://go-review.googlesource.com/c/go/+/362794
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Katie Hockman <katie@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Roland Shoemaker <roland@golang.org>
Robert Griesemer [Thu, 11 Nov 2021 19:57:43 +0000 (11:57 -0800)]
go/types, types2: document nil scope for imported and instantiated Func objects
Also, don't set the scope anymore when instantiating (substituting)
a signature.
Per discussion with rfindley.
Change-Id: I560d4571c7ff14b0df3e15fece634cb5f9f94a99
Reviewed-on: https://go-review.googlesource.com/c/go/+/363435
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>
Keith Randall [Thu, 11 Nov 2021 16:45:02 +0000 (08:45 -0800)]
cmd/compile: ensure stenciled function bodies are nonempty
Our compiler gets confused between functions that were declared
with no body, and those which have a body but it is empty.
Ensure that when stenciling, we generate a nonempty body.
The particular test that causes this problem is in
cmd/compile/internal/gc/main.go:enqueueFunc. It thinks that if
a function has no body, then we need to generate ABI wrappers for
it, but not compile it.
Fixes #49524
Change-Id: Id962666a2098f60a2421484b6a776eafdc4f4a63
Reviewed-on: https://go-review.googlesource.com/c/go/+/363395
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Dan Scales <danscales@google.com>
Ian Lance Taylor [Thu, 11 Nov 2021 04:28:45 +0000 (20:28 -0800)]
go/build: skip rune literals when looking for go:embed
Fixes #49514
Change-Id: Id687eead731ba49974f11d2e5b489f11eff7d07b
Reviewed-on: https://go-review.googlesource.com/c/go/+/363275
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Michael Anthony Knyszek [Thu, 11 Nov 2021 17:31:36 +0000 (17:31 +0000)]
runtime: adjust TestPhysicalMemoryUtilization to handle large page sizes
Currently TestPhysicalMemoryUtilization can fail on systems with large
physical page sizes like 64 KiB because all the of the holes to be
scavenged are not aligned to the page size. The holes themselves are 64
KiB so this is actually quite likely.
Bump the size of the allocations for systems with larger physical page
sizes, and add additional slack to the threshold for unaligned pieces of
the holes that may be unaligned.
Fixes #49411.
Change-Id: Iafb35b8761dc9cdc53d3745c4771b1a64c5c97b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/363415
Trust: Michael Knyszek <mknyszek@google.com> Reviewed-by: David Chase <drchase@google.com>
Michael Anthony Knyszek [Wed, 10 Nov 2021 22:03:28 +0000 (22:03 +0000)]
runtime/debug: make TestFreeOSMemory more robust
FreeOSMemory relies on the function FreeOSMemory increasing HeapReleased
as opposed to the background scavenger, because it reads memory stats
*after* the free of a large allocation. However, before that even
happens, the background scavenger can swoop in and release all that
memory, making it appear as if FreeOSMemory didn't do anything.
This change modifies the test to just make sure that the large
allocation's memory is returned to the OS *somehow*, by the end of the
test. It doesn't really care which happens. It also increases the size
of that large allocation to increase the likelihood that the test isn't
relying 100% on the background scavenger, and that FreeOSMemory is doing
some of the work.
Fixes #49478.
Change-Id: Ief1d839753720ebb88cbb616c46302293ee2d19c
Reviewed-on: https://go-review.googlesource.com/c/go/+/363414 Reviewed-by: David Chase <drchase@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Pavel [Mon, 8 Nov 2021 14:29:16 +0000 (14:29 +0000)]
database/sql: prevent closes slices from assigning to free conn
In function connectionCleanerRunLocked append to closing slice affects db.freeConns and vise versa. Sometimes valid connections are closed and some invalid not.
Change-Id: I5282f15be3e549533b7d994b17b2060db3c0e7da
GitHub-Last-Rev: b3eb3ab6f49c036519f777fc7189e9507010c166
GitHub-Pull-Request: golang/go#49429
Reviewed-on: https://go-review.googlesource.com/c/go/+/362214 Reviewed-by: Daniel Theophanes <kardianos@gmail.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Robert Griesemer [Thu, 11 Nov 2021 16:36:15 +0000 (08:36 -0800)]
spec: fix a broken link
Thanks for jtagcat@ for finding this.
Change-Id: If7324808edbae19ec8bf503b04e0426f3fb3b47a
Reviewed-on: https://go-review.googlesource.com/c/go/+/363394
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
hasheddan [Thu, 11 Nov 2021 15:02:13 +0000 (10:02 -0500)]
obj/riscv: fix link to risc-v dwarf register numbers
The repository name and structure in the RISC-V GitHub org has been
modified, rendering the existing link invalid. This updates to point at
the new location of the RISC-V DWARF specification.
Change occured in https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/208
jiahua wang [Sat, 2 Oct 2021 14:50:31 +0000 (22:50 +0800)]
io: add error check to TeeReader Example
Change-Id: I0b94bdced47483c6412e9979ce2d103fbfc52afb
Reviewed-on: https://go-review.googlesource.com/c/go/+/353729
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Carlos Amedee <carlos@golang.org>
Cuong Manh Le [Thu, 11 Nov 2021 08:29:38 +0000 (15:29 +0700)]
cmd/compile: fix missing ddd when building call for function instantiation closure
When building a call expression for function instantiation closure, if
it's a variadic function, the CallExpr.IsDDD must be set for typecheck
to work properly. Otherwise, there will be a mismatch between the
arguments type and the function signature.
Fixes #49516
Change-Id: I0af90ee3fcc3e6c8bba8b20e331e044cbce17985
Reviewed-on: https://go-review.googlesource.com/c/go/+/363314
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Cherry Mui [Wed, 10 Nov 2021 00:50:47 +0000 (19:50 -0500)]
runtime: bypass scheduler when doing traceback for goroutine profile
When acquire a goroutine profile, we stop the world then acquire a
stack trace for each goroutine. When cgo traceback is used, the
traceback code may call the cgo traceback function using cgocall.
As the world is stopped, cgocall will be blocked at exitsyscall,
causing a deadlock. Bypass the scheduler (using asmcgocall) to fix
this.
Change-Id: Ic4e596adc3711310b6a983d73786d697ef15dd72
Reviewed-on: https://go-review.googlesource.com/c/go/+/362757
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Than McIntosh [Thu, 11 Nov 2021 12:06:24 +0000 (07:06 -0500)]
net: skip new testpoint TestUDPIPVersionReadMsg on plan9
Skip TestUDPIPVersionReadMsg on plan9, since it does things
not supported on that OS.
Change-Id: Icd1716fb5ed4e8877e57acb8c851ec3be72e83e2
Reviewed-on: https://go-review.googlesource.com/c/go/+/363354 Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Than McIntosh [Wed, 10 Nov 2021 20:36:25 +0000 (15:36 -0500)]
cmd/compile: use canonical stringslice/ representations in abiutils
A chunk of code in abiutils was synthesizing the internals of a Go
string type as "struct { unsafe.Pointer, uintptr }" instead of the
more canonical representation "struct { *uint8, int }" used elsewhere
in the compiler. The abiutils type was being pulled into the code
during late call expansion, which resulted in two different entries in
the SSA named value table for the same variable piece, each with
different types; this then confused DWARF location list generation.
This patch changes the abiutils synthesized type to be consistent with
other parts of the back end, and makes a similar change for
synthesized slice types (use "struct { *uint8, int, int }").
Fixes #47354.
Change-Id: If789031cdc7abaf215bc75ee6eb863defbe530be
Reviewed-on: https://go-review.googlesource.com/c/go/+/362715
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Mui <cherryyz@google.com>
Than McIntosh [Wed, 10 Nov 2021 15:44:00 +0000 (10:44 -0500)]
cmd/compile: include register-resident output params in DWARF-gen
During the register ABI work, a change was made in CL 302071 to
"stackframe" to treat register-resident output parameter (PARAMOUT)
variables that same as locals, which meant that if they were unused,
we'd delete them from the "Dcl" slice. This has the effect of making
them invisible to DWARF generation later on in the pipeline, meaning
that we don't get DIEs for them in the debug info. This patch fixes
the problem by capturing these params prior to optimization and then
adding them back in for consideration when we're processing the
params/locals of a function during DWARF generation.
Fixes #48573.
Change-Id: I2b32882911c18f91c3e3d009486517522d262685
Reviewed-on: https://go-review.googlesource.com/c/go/+/362618
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Than McIntosh [Mon, 8 Nov 2021 18:53:55 +0000 (13:53 -0500)]
cmd/compile/internal/ssa: fix debug location gen issue with zero width ops
Revamp the way that buildLocationLists() handles zero-width
operations, to fix a couple of problems that result in bad debug
locations.
The problematic scenario in this specific bug is where you have a
parameter arriving in a register X, then a spill of register X to
memory as the first non-zero-width instruction in the function.
Example:
The existing buildLocationLists implementation effectively buffers or
bundles changes from zero-width instructions until it it sees a
non-zero-width instruction, but doing that in this case winds up
making it look as though the parameter is live into the function in
memory, not in a register.
The fix for this to separate out zero-width ops into two distinct
categories: those that whose lifetimes begin at block start (ex:
OpArg, Phi) and those whose effects are taking place at the nearest
non-zero-width instruction (ex: OpSelect0). In this patch we now
handle the first category of ops in an initial pre-pass for each
block, and leave the second category for the main pass through the
block. See the notes on the issue below for a more detailed
explanation of the failure mode.
Fixes #46845.
Change-Id: I27488d4c041019d5a0b897b7cf53000f63aab1cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/362244
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Jason A. Donenfeld [Wed, 10 Nov 2021 00:04:12 +0000 (01:04 +0100)]
net: add conversion from AddrPort to TCPAddr to complement existing inverse
We already have various member functions of TCPAddr that return an
AddrPort, but we don't have a helper function to go from a AddrPort to a
TCPAddr. UDP has this, but it was left out of TCP. This commit adds the
corresponding function.
Updates #49298.
Change-Id: I85732cf34f47c792fe13a6b4af64fd4b0e85d06a
Reviewed-on: https://go-review.googlesource.com/c/go/+/362596
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Dmitry Vyukov [Wed, 10 Nov 2021 18:51:34 +0000 (19:51 +0100)]
cmd/link/internal/loadelf: better error message for ignored symbols
Currently it's quite hard to debug these error messages about ignored symbols
because there are only some numbers and no symbol name. Add symbol name. Before:
135029: sym#952: ignoring symbol in section 11 (type 0)
After:
135029: sym#952 (_ZN11__sanitizer9SpinMutexC5Ev): ignoring symbol in section 11 (type 0)
Change-Id: I7fec50b5798068c74827376613be529803838c5a
Reviewed-on: https://go-review.googlesource.com/c/go/+/363034
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitry Vyukov <dvyukov@google.com>
Ian Lance Taylor [Wed, 10 Nov 2021 00:17:50 +0000 (16:17 -0800)]
lib/time, time/tzdata: update to 2021e
Doing this a little early in the release cycle as there have been some
changes in the handling of old timezones. They should continue to
work as expected, but more testing time may be useful.
For #22487
Change-Id: I3686fed79a052c46112445055044cff5842f2a45
Reviewed-on: https://go-review.googlesource.com/c/go/+/362874
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>
Ian Lance Taylor [Wed, 10 Nov 2021 17:35:59 +0000 (09:35 -0800)]
misc/cgo/testshared: correct test of gccgo version number
We still don't run the gccgo tests, because they don't run in module mode.
But now we at least get the version number check right.
Change-Id: Ifde4512c30605d1cb7e3a521f381a05c783549b7
Reviewed-on: https://go-review.googlesource.com/c/go/+/362996
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>
Changkun Ou [Mon, 26 Jul 2021 08:56:30 +0000 (10:56 +0200)]
sync: clarify the validity to call Map methods inside Range
This change clarifies that calling all Map methods inside the callback
of Range is allowed. For further assurance, a nested range call test
is also added.
Fixes #46399
Change-Id: I0a766a5c1470e6b573ec35df1ccd62b2e46f1561
Reviewed-on: https://go-review.googlesource.com/c/go/+/337389 Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Thu, 11 Nov 2021 00:06:18 +0000 (16:06 -0800)]
test: add test that was miscompiled by gccgo
For #49512
Change-Id: Ic08652a4ec611b27150bf10b1118c1395715e5d0
Reviewed-on: https://go-review.googlesource.com/c/go/+/363156
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Than McIntosh <thanm@google.com>
Jason A. Donenfeld [Fri, 5 Nov 2021 12:02:09 +0000 (13:02 +0100)]
net: do more faithful conversion from AddrPort to UDPAddr
A UDPAddr with a nil IP is a valid state, representing an AF-agnostic
unspecified address, so checking for addr.IsValid() isn't correct;
remove that, as it's only needed in the UDP rx path where it can be
added. Secondly, forcing everything to be IPv6 also is not correct, and
was likely done when the missing .AsSlice() made doing the right thing
less ergonomic. Fix this by using .AsSlice(), which properly preserves
IP version.
Change-Id: Idd1eaecd4076f32a843f859a0a9802ef98f956d3
Reviewed-on: https://go-review.googlesource.com/c/go/+/361478
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Robert Griesemer [Wed, 10 Nov 2021 21:36:23 +0000 (13:36 -0800)]
cmd/compile/internal/types2: remove structuralString in favor of inlined code
structuralString was used only in one place (for built-in copy).
Remove it in favor of custom and more efficient inlined code.
Follow-up on feedback received for CL 363075.
Change-Id: Ic5857c47255c5c712be7971aae4542fef9960fe6
Reviewed-on: https://go-review.googlesource.com/c/go/+/363154
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 [Wed, 10 Nov 2021 20:11:03 +0000 (12:11 -0800)]
cmd/compile/internal/types2: slightly relax notion of structural type
If we have all channel types in a constraint, there is no structural type
if they don't all have the same channel direction (and identical element
types, of course). By allowing different channel types for the purposes of
the structural type, as long as there is not a send-only _and_ a receive-
only channel in the type set, we make it possible to find a useful, if
restricted by channel direction, structural type where before there was
none.
So if we have unrestricted and send-only channels, the structural type is
the send-only channel, and vice versa.
For all operations on channels that rely on a structural type, it's always
ok to have an unrestricted channel, so this is not affecting their behavior.
But it makes those operations more flexible in the presence of type parameters
containing mixed channel types.
For constraint type inference, where we currently may not infer a channel
at all, this change allows us to infer a more restricted channel (send- or
receive-only). If the inferred channel type is a valid type argument we win;
if not we haven't lost anything.
Use structuralType for send and receive operations and adjust related
error messages (the error message that change are the ones involving
type parameters, so historic error messages are preserved).
Fixes #45920.
Change-Id: If3a64d29c37e7734d3163df330f8b02dd032bc60
Reviewed-on: https://go-review.googlesource.com/c/go/+/363075
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
Dan Scales [Wed, 10 Nov 2021 19:39:18 +0000 (11:39 -0800)]
cmd/compile: don't do Resolve on OKEY identifiers during import
For generic functions, we can export untransformed OKEY nodes, and the
key identifier is written as an ONONAME. But in this case, we do not
want to call Resolve() on the identifier, since we may resolve to a
global type (as happens in this issue) or other global symbol with the
same name, if it exists. We just want to keep the key identifier as an
Ident node.
To solve this, I added an extra bool when exporting an ONONAME entry,
which indicates if this entry is for a key or for a global (external)
symbol. When the bool is true (this is for a key), we avoid calling
Resolve().
Fixes #49497
Change-Id: Ic8fa93d37bcad2110e0e0d060080b733e07e35d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/363074
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Adds a new flag to 'go mod vendor' which overrides the default
'vendor' destination directory. This can be helpful for writing the
vendor tree to a temporary location for use by other tools.
The argument can be a relative or an absolute path.
This flag has no other influence on how the command behaves.
Fixes #47327
Change-Id: I4502931127616b181dc90a2066d2fb57bfe48f96
Reviewed-on: https://go-review.googlesource.com/c/go/+/338149 Reviewed-by: Bryan C. Mills <bcmills@google.com> Reviewed-by: Jay Conrod <jayconrod@google.com>
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Robert Griesemer [Wed, 10 Nov 2021 16:33:26 +0000 (08:33 -0800)]
cmd/compile/internal/types2: move some functions into different files (cleanup)
- move structuralType/structuralString into type.go
- move functions exported for the compiler into compilersupport.go
- updated/added comments
- removed AsNamed and AsInterface - not needed by compiler
No semantic changes.
Change-Id: Ia454a49edafd627c2a25b0b71db4aa93ddd7f1f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/362995
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>
Robert Findley [Wed, 10 Nov 2021 16:21:57 +0000 (11:21 -0500)]
go/doc: don't treat functions returning type parameters as constructors
Functions returning type parameters were erroneously being interpreted
as 'constructors' of their type parameter, resulting in them being
excluded from documentation. Fix this by explicitly excluding type
parameters when looking for defined type names among function results.
Fixes #49477
Change-Id: I22510f655f47e192a852332df5b91740f46c51eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/362758
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com> Reviewed-by: Jonathan Amsterdam <jba@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Robert Griesemer [Wed, 10 Nov 2021 16:12:21 +0000 (08:12 -0800)]
cmd/compile/internal/types2: rename structure to structuralType
And rename structureString to structuralString.
Now that we have an updated definition for structural types in
the (forthcoming) spec, name the corresponding function accordingly.
No semantic changes.
Change-Id: Iab838f01a37075bedf2d8bc4f166b0217672b85f
Reviewed-on: https://go-review.googlesource.com/c/go/+/362994
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 [Wed, 10 Nov 2021 05:12:03 +0000 (21:12 -0800)]
cmd/compile/internal/types2: use type variables consistently in Checker.conversion
We have V and T and Vu and Tu. When calling the various isX predicates
consistently use Vu and Tu.
(We could also use V an T because the predicates call under anyway,
but using Vu and Tu removes an unnecessary call to Named.under if
V or T are *Named.)
Also, removed some outdated comments.
Change-Id: I6fcd9ce5f6292e89ac2afd597b72fd0790e84ff1
Reviewed-on: https://go-review.googlesource.com/c/go/+/362895
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>
Robert Griesemer [Wed, 10 Nov 2021 04:22:26 +0000 (20:22 -0800)]
doc: make a copy of the latest Go 1.17 spec
This will allow us to compare the changes made for Go 1.18.
Change-Id: I1456270b201967f5cb05e66cec556939e6e33265
Reviewed-on: https://go-review.googlesource.com/c/go/+/362894
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Hajime Hoshi [Fri, 5 Nov 2021 14:50:55 +0000 (23:50 +0900)]
runtime: fix unworkable comments for go:nosplit
Change-Id: I71c29a2dc7e5b2b6bc35093535228d2907b16b47
Reviewed-on: https://go-review.googlesource.com/c/go/+/361595
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Hajime Hoshi <hajimehoshi@gmail.com>
Robert Findley [Sun, 31 Oct 2021 15:33:00 +0000 (11:33 -0400)]
go/parser: allow parsing aliases with type parameters
We already guard against this in the type checker, and it will
eventually be allowed per the accepted proposal.
Add a placeholder error code for the corresponding type checker error.
Change-Id: I5cc2f1413ecc89ec2094f7178fdb156fb8cc2e43
Reviewed-on: https://go-review.googlesource.com/c/go/+/360235
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 23:48:16 +0000 (18:48 -0500)]
go/types: disallow type cycles through type parameter lists
This is a port of CL 361922 to go/types.
Change-Id: I790c8121a640c25fb655c926fb434d667dd59f76
Reviewed-on: https://go-review.googlesource.com/c/go/+/362756
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Griesemer [Tue, 9 Nov 2021 00:09:11 +0000 (16:09 -0800)]
cmd/compile/internal/types2: disallow type cycles through type parameter lists
If we reach a generic type that is part of a cycle
and we are in a type parameter list, we have a cycle
through a type parameter list, which is invalid.
Fixes #49439.
Change-Id: Ia6cf97e1748ca0c0e61c02841202050091365b0b
Reviewed-on: https://go-review.googlesource.com/c/go/+/361922
Trust: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
Robert Griesemer [Tue, 9 Nov 2021 04:17:22 +0000 (20:17 -0800)]
cmd/compile/internal/types2: rename Checker.cycle to Checker.validCycle
Also, invert the boolean result. This matches Checker.validType;
it's also easier to understand. Preparation for the next CL which
detects cycles through type parameter lists.
Change-Id: I00a75d2359ca20827c9bf406945508716c826fc4
Reviewed-on: https://go-review.googlesource.com/c/go/+/362336
Trust: Robert Griesemer <gri@golang.org> Reviewed-by: Robert Findley <rfindley@google.com>
Robert Findley [Tue, 9 Nov 2021 23:40:38 +0000 (18:40 -0500)]
go/types: rename Checker.cycle to Checker.validCycle
This is a clean port of CL 362336 to go/types.
Change-Id: Iafeae7024fbb2872b07748affcea9676324ea59e
Reviewed-on: https://go-review.googlesource.com/c/go/+/362755
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 21:41:01 +0000 (16:41 -0500)]
go/types: remove most asX converters (cleanup)
This is a port of CL 362118 to go/types, which is itself a roll-forward
of CL 362254, containing a bugfix.
Change-Id: I20067c7adf56bf64fe9ad080d998a7aefbdc1053
Reviewed-on: https://go-review.googlesource.com/c/go/+/362617
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 19:20:11 +0000 (14:20 -0500)]
go/types: report error for incomplete struct composite literal type
This is a port of CL 361412 to go/types.
Change-Id: Ie5bccc7faba7ca9230e712f867b27ca9dcddba79
Reviewed-on: https://go-review.googlesource.com/c/go/+/362739
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 19:56:39 +0000 (14:56 -0500)]
go/types: don't return an array type with invalid length
In preparation for porting CL 361412, fix a discrepancy in go/types,
where [-1]T is returned for an array type with invalid length.
Change-Id: Ia32f5b66c9c561ccf0c32af1922fc4690c66dbc3
Reviewed-on: https://go-review.googlesource.com/c/go/+/362738
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Michael Matloob [Tue, 19 Oct 2021 18:05:29 +0000 (14:05 -0400)]
cmd/go: add workspace pruning mode
[ this is a roll-forward of golang.org/cl/357169 with minor changes
to fix the cmd/go/internal/modload tests: because they don't run the
go command, some initialization isn't run on the test and modroots is
empty in cases it can't be when the full command setup is done. So
directly check for workFilePath != "" instead of calling inWorkspaceMode
which checks that Init is called first, and check that modRoots is non
empty when calling mustGetSingleMainModule.]
This change corrects a bug in the handling of module loading of
workspaces. Namely, there is an assumption by the module pruning code
that if a root module is selected then the packages of that module can
be resolved without loading the whole module graph. This is not true
in workspace mode because two workspace modules can require different
versions of a dependency. Worse, one workspace module can directly
require a depencency that is transitively required by another
workspace module, changing the version of that module loaded in the
fully expanded graph.
To correct this, a new 'workspace' pruning mode is added where the
roots are the workspace modules themselves, satisfying the assumption
made by the module pruning logic.
The rest of this change accounts for the new pruning mode where it's
used and correctly sets the requirements in this pruning mode.
Change-Id: I8bdf4b30f669c1ded0ed8a5dd202ac8d1939bbbd
Reviewed-on: https://go-review.googlesource.com/c/go/+/362754
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Manlio Perillo [Sun, 16 May 2021 17:52:37 +0000 (19:52 +0200)]
cmd/dist: wait for exit in bgwait after a fatal error
Currently, when a command scheduled by bgrun fails, bgwait returns to
the caller even in case the fatal error was detected.
In case of a syntax error in one of the standard packages, as an
example, the runInstall function will try to read the generated archive
file, only to fail since the file does not exist. Since the runInstall
function is called in a goroutine, cmd/dist will continue to report
errors until all background goroutines are done.
Update the bgwait function to wait until program termination in case of
a fatal error, since returning to the caller (with an error, as an
example) will cause cmd/dist to report an additional error during the
next build phase.
Fixes #45410
Change-Id: If89976abad70f8d6ec79b2a5a1f2306e9c034c5a
Reviewed-on: https://go-review.googlesource.com/c/go/+/320311
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Keith Randall [Tue, 12 Oct 2021 23:50:35 +0000 (16:50 -0700)]
encoding/xml: add generic encoding test
Fixes #48521
Change-Id: Id8402bcff243c0ab19e4ec0b138b9af8c111f88d
Reviewed-on: https://go-review.googlesource.com/c/go/+/355492
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Dan Scales <danscales@google.com>
Robert Findley [Tue, 9 Nov 2021 19:10:37 +0000 (14:10 -0500)]
go/types: clearer object string for type parameters
This is a port of CL 361401 to go/types.
Change-Id: I5b1c7cf1d7a819b2902c304f884492ec02c7eaa1
Reviewed-on: https://go-review.googlesource.com/c/go/+/362737
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
David Chase [Tue, 9 Nov 2021 16:47:49 +0000 (11:47 -0500)]
cmd/compile: add line number test for #49436
This enhances the existing line number test
to allow a specific -gcflags (e.g., -G=3)
and to permit ignoring duplicate line numbers
(which is arguably a bug, but not THIS bug,
and it lowers the risk of a flaky test).
Limited to Linux/Darwin and amd64/arm64,
also tests with "unified" mangling.
And, using these new powers, adds a test.
Updates #49436.
Change-Id: I09c82e6a08d53edd5a752522a827e872d3e16e0b
Reviewed-on: https://go-review.googlesource.com/c/go/+/362714
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Dmitri Shuralyov [Tue, 9 Nov 2021 18:54:57 +0000 (13:54 -0500)]
all: update vendored golang.org/x/tools for Go 1.18 release
The Go 1.18 code freeze has recently started. This is a time to update
all golang.org/x/... module versions that contribute packages to the
std and cmd modules in the standard library to latest master versions.
This CL updates only the tools module, keeping mod unchanged because
its lastest commit isn't ready to be vendored yet.
For #36905.
Updates #49350.
Change-Id: Ib39713d28a55fc9ec79058aab9919eba912def5f
Reviewed-on: https://go-review.googlesource.com/c/go/+/361094
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Dmitri Shuralyov [Tue, 9 Nov 2021 18:49:45 +0000 (13:49 -0500)]
all: update vendored golang.org/x/crypto for Go 1.18 release
The Go 1.18 code freeze has recently started. This is a time to update
all golang.org/x/... module versions that contribute packages to the
std and cmd modules in the standard library to latest master versions.
This CL updates only the crypto module, as well as the TestDependencies
policy to accommodate the dependency order change done in CL 345649.
The next CL will update further modules.
For #36905.
Change-Id: If93160d8d72ca86e9995aaf5bdfc3d2c20b4695d
Reviewed-on: https://go-review.googlesource.com/c/go/+/362736
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com>
Dmitri Shuralyov [Tue, 9 Nov 2021 18:31:45 +0000 (13:31 -0500)]
all: update vendored golang.org/x/{net,text} for Go 1.18 release
The Go 1.18 code freeze has recently started. This is a time to update
all golang.org/x/... module versions that contribute packages to the
std and cmd modules in the standard library to latest master versions.
This CL updates only the net, text modules.
The next CL will update further ones.
For #36905.
Change-Id: I9a5ac3cca22da961cfd09f3202e01e1187d42bdd
Reviewed-on: https://go-review.googlesource.com/c/go/+/362735
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com>
Dmitri Shuralyov [Tue, 9 Nov 2021 18:31:45 +0000 (13:31 -0500)]
all: update vendored golang.org/x/{arch,sys,term} for Go 1.18 release
The Go 1.18 code freeze has recently started. This is a time to update
all golang.org/x/... module versions that contribute packages to the
std and cmd modules in the standard library to latest master versions.
This CL updates only the lower-level modules arch, sys, term for better
bisection. The next CL will update further ones.
For #36905.
Change-Id: I455428c051ec49b446b8b558a6f579cd9be4d796
Reviewed-on: https://go-review.googlesource.com/c/go/+/362734
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Ian Lance Taylor [Tue, 9 Nov 2021 02:30:30 +0000 (18:30 -0800)]
os: clarify that File.{Read,Write} use the buffer
Fixes #49470
Change-Id: I81fd4b0e2eef1d8d430b5d1d10c4f824e803a75c
Reviewed-on: https://go-review.googlesource.com/c/go/+/362335
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Ian Lance Taylor [Mon, 8 Nov 2021 21:57:50 +0000 (13:57 -0800)]
bufio: document that NewWriter can return its argument
Fixes #49446
Change-Id: Ib0b53a7dd5d567a2dd0bdf29f53d276587b60afb
Reviewed-on: https://go-review.googlesource.com/c/go/+/361921
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>
Chaoqun Han [Mon, 8 Nov 2021 14:58:51 +0000 (22:58 +0800)]
cmd/compile: NewSelectorExpr use n.Pos instead of base.Pos
Fixes #49436
Change-Id: I4c8851e7aaee631d5eb22f2ef0aea5a25e936d87
Reviewed-on: https://go-review.googlesource.com/c/go/+/361917
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Trust: David Chase <drchase@google.com>
Robert Findley [Tue, 9 Nov 2021 16:19:20 +0000 (11:19 -0500)]
go/types: make object test an external test
This is a port of CL 361409 to go/types.
Change-Id: I17ccf8a5b4ba715fd8a87ea2c1811700fb1157e3
Reviewed-on: https://go-review.googlesource.com/c/go/+/362538
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 16:14:37 +0000 (11:14 -0500)]
go/types: check non-generic conversions first
This is a clean port of CL 361269 to go/types.
Change-Id: I2caaf08eabdf1707ae83ec1e628fd26f21b2b8e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/362616
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Robert Findley [Tue, 9 Nov 2021 16:09:13 +0000 (11:09 -0500)]
go/types: minor cleanups in predicates.go
This is a clean port of CL 360956 to go/types.
Change-Id: Iac437e72bb760e7e90236a86e7473d6a440df081
Reviewed-on: https://go-review.googlesource.com/c/go/+/362615
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Since GCC version 11, the 64-bit version of GCC starting files are
now suffixed by "_64" instead of being stored without suffix under
"ppc64" multilib directory.
Change-Id: Ibe53521ed24d36e5f6282e3574849b9ae11a1e9a
Reviewed-on: https://go-review.googlesource.com/c/go/+/362594 Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Ian Lance Taylor <iant@golang.org>