We delay all transformations on generic functions, and only do them on
instantiated functions, for several reasons, of which one is that
otherwise the compiler won't understand the relationship between
constrained type parameters. In an instantiation with shape arguments,
the underlying relationship between the type arguments are clear and
don't lead to compiler errors.
This issue is because I missed delaying assignment transformations for
variable declarations. So, we were trying to transform an assignment,
and the compiler doesn't understand the relationship between the T and U
type parameters.
The fix is to delay assignment transformations for variable declarations
of generic functions, just as we do already for normal assignment
statements.
A work-around for this issue would be to just separate the assignment
from the variable declaration in the generic function (for this case of
an assignment involving both of the constrained type parameters).
Fixes#50147
Change-Id: Icdbcda147e5c4b386e4715811761cbe73d0d837e
Reviewed-on: https://go-review.googlesource.com/c/go/+/371534
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
We can import an shape-instantiated function/method for inlining
purposes. If we are instantiating the methods of a instantiated type
that we have seen, and it happens to need a shape instantiation that we
have imported, then don't re-create the instantiation, since we will end
up with conflicting/duplicate definitions for the instantiation symbol.
Instead, we can just use the existing imported instantation, and enter
it in the instInfoMap[].
Fixes#50121
Change-Id: I6eeb8786faad71106e261e113048b579afad04fa
Reviewed-on: https://go-review.googlesource.com/c/go/+/371414
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
A "RET f(SB)" wasn't assembled correctly in a leaf function with
non-zero frame size. Follows CL 371034, for MIPS(32/64)(be/le)
and S390X. Other architectures seem to do it right. Add a test.
Change-Id: I41349a7ae9862b924f3a3de2bcb55b782061ce21
Reviewed-on: https://go-review.googlesource.com/c/go/+/371214
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
h2_bundle.go is automatically generated from x/net/http2. Any
formatting changes within that file need to be first made upstream.
This brings the contents of h2_bundle.go back in line with the
upstream generator, fixing the cmd/internal/moddeps test that is
currently failing on the longtest builders.
For #49884
Change-Id: I5757240b77e250e0026b8a52a0e867e1578ec2d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/371297
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Fix up a unit test to make it more friendly for coverage runs.
Currently on tip if you do
cd ${GOROOT}/src ; go test -cover cmd/compile/...
it will cause a failure in the TestGoAMD64v1 testpoint of
cmd/compile/internal/amd64, the reason being that this testpoint
copies and reruns the test executable, expecting the rerun to produce
only the output "PASS", whereas if "-cover" is used, the output will
include percentage of statements covered as well. To fix, rework the
test to tolerate additional output if coverage is enabled.
Change-Id: I2512e06ca06e5f38108f2891ff84276d148c4f9e
Reviewed-on: https://go-review.googlesource.com/c/go/+/371234
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Although most of the code seems to be already implemented
to support this for general use, it didn't make it in for
Go 1.18, so for now we should at least document that it's
only for use by the go command.
Change-Id: Id559e72d590aedeaaa50bcf880bca1a385d858dd
Reviewed-on: https://go-review.googlesource.com/c/go/+/370954
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Katie Hockman <katie@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
And then revert the bootstrap cmd directories and certain testdata.
And adjust tests as needed.
Not reverting the changes in std that are bootstrapped,
because some of those changes would appear in API docs,
and we want to use any consistently.
Instead, rewrite 'any' to 'interface{}' in cmd/dist for those directories
when preparing the bootstrap copy.
A few files changed as a result of running gofmt -w
not because of interface{} -> any but because they
hadn't been updated for the new //go:build lines.
Fixes#49884.
Change-Id: Ie8045cba995f65bd79c694ec77a1b3d1fe01bb09
Reviewed-on: https://go-review.googlesource.com/c/go/+/368254
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
As written, the conversion P(x), where P and the type
of x are type parameters with identical underlying types
(i.e., identical constraints), is valid. However, unless
the type of x and P are identical (which is covered with
the assignability rule), such a conversion is not valid
in general (consider the case where both type parameters
are different type parameters with constraint "any").
This change adjusts the rules to prohibit type parameters
in this case. The same reasoning applies and the analogue
change is made for pointer types.
The type checker already implements these updated rules.
Change-Id: Id90187900cb2820f6a0a0cf582cf26cdf8addbce
Reviewed-on: https://go-review.googlesource.com/c/go/+/371074
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The test cases for this test had listed specific errors, but the
specific error values were ignored in favor of just calling
isDeadlineExceeded.
Moreover, ENOBUFS errors (which can legitimately occur in the test if
the network interface also happens to be saturated when the timeout
occurs) were not handled at all.
Now the test relies only on the timeout: we iterate until we have seen
two of the expected timeout errors, and if we see ENOBUFS instead of
"deadline exceeded" we back off to give the queues time to drain.
Fixes#49930
Change-Id: I258a6d5c935d9635b02dffd79e197ba9caf83ac8
Reviewed-on: https://go-review.googlesource.com/c/go/+/370882
Trust: Bryan Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This change applies the same transformation as in CL 366774,
but to the net package.
testUnixAddr was using os.CreateTemp to obtain a unique socket path,
but then calling os.Remove on that path immediately. Since the
existence of the file is what guarantees its uniqueness, that could
occasionally result in testUnixAddr returning the same path for two
calls, causing the tests using those paths to fail — especially if
they are the same test or are run in parallel.
Instead, we now create a unique, short temp directory for each call,
and use a path within that directory for the socket address.
For #34611
Change-Id: I8e13b606abce2479a0305f7aeecf5d54c449a032
Reviewed-on: https://go-review.googlesource.com/c/go/+/370694
Trust: Bryan Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Passing in an explicit testing.TB gives two benefits:
1. It allows the helper to fail the test itself, instead of returning
an error to the caller. A non-nil error invariably fails the
calling test, and none of these callers bother to add detail to the
error when logging it anyway so returning the error just added
noise to the test bodies.
2. It allows the helper to use t.Cleanup to perform any needed cleanup
tasks, which will be used in CL 370695 to clean up temp directories
used as namespaces for unix socket paths.
For #34611
Change-Id: I805e701687c12de2caca955649369294229c10b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/370696
Trust: Bryan Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
In some leaf functions using "RET foo(SB)", the jump may be incorrectly
translated into "JMP LR" instead of "JMP foo(SB)".
Such is the case when compiling the autogenerated function in k8s
k8s.io/kubernetes/pkg/kubelet/server/stats.(*resourceAnalyzer).GetPodVolumeStats.
Fixes#50048
Change-Id: Icdf65fcda6b3c5eb9d3ad5c7aad04add9419dffb
Reviewed-on: https://go-review.googlesource.com/c/go/+/371034
Run-TryBot: Paul Murphy <murp@ibm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Trust: Paul Murphy <murp@ibm.com>
The exact error isn't actually relevant to the test,
and may depend on whether the Go or cgo resolver is used.
Also run the test in parallel, because it spends most of its time
sleeping in between lookups.
Fixes#38767Fixes#43140
Change-Id: I2d64ffddf2eb114a69ed3242daa9a9e4a5679f67
Reviewed-on: https://go-review.googlesource.com/c/go/+/369037
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TestUnixAndUnixpacketServer deferred a call to os.Remove on the local
address of a dialed unix domain socket, in an attempt to remove the
socket from the server. However, that call appears to be neither
necessary nor correct.
In this test, the file that needs to be unlinked is the one attached
to the listener — but the listener's Close method already does that
(see the Unlink call in (*UnixListener).close), so there is no need
for the test itself to do the same.
Moreover, the local address is not something that is sensible to
delete — on Linux, it is empirically always the literal string "@" —
and the Addr returned by c.LocalAddr is not reliably non-nil on all
platforms (see #34611).
Since we don't need to do anything with the local address, we shouldn't.
At best, this is a benign Remove of a file that doesn't exist anyway;
at worst, it is a nil-panic.
Fixes#34611
Change-Id: Ie072b3388d884d60e819d1df210fa7d3e2eed124
Reviewed-on: https://go-review.googlesource.com/c/go/+/370695
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TestListenerClose had been asserting that a Dial to the newly-closed
address always fails, on the assumption that the listener's address
and port would not be reused by some other listener that could then
accept the connection.
As far as I can tell, that assumption is not valid: the Dial after
Close may well connect to a Listener opened for some other test, or
even one opened by a completely different process running concurrently
on the same machine.
Fixes#38700
Change-Id: I925ed1b2ccb556135a2c5be0240d1789ed27d5fc
Reviewed-on: https://go-review.googlesource.com/c/go/+/370666
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
In identical(), we don't want any to match a shape empty-interface type
for the identStrict option, since IdenticalStrict() is specifically not
supposed to match a shape type with a non-shape type.
There is similar code in (*Type).cmp() (TINTER case), but I don't
believe that we want to disqualify shape types from matching any in this
case, since cmp() is used for back-end code, where we don't care about
shape types vs non-shape types.
The issue mainly comes about when 'any' is used as a type argument
(rather than 'interface{}'), but only with some complicated
circumstances, as shown by the test case. (Couldn't reproduce with
simpler test cases.)
Fixes#50109
Change-Id: I3f2f88be158f9ad09273237e1d346bc56aac099f
Reviewed-on: https://go-review.googlesource.com/c/go/+/371154
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Currently, File.Close only documents that "an" error will be returned
by pending I/O operations. Update the documentation to say that error
is specifically ErrClosed.
Change-Id: Ica817c9196ad6cb570c826789d37a4ff15a5d13d
Reviewed-on: https://go-review.googlesource.com/c/go/+/371015
Trust: Austin Clements <austin@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This sometimes times out and we don't have any useful output for
debugging it. Hopefully this will help.
For #37405.
Change-Id: I79074e6fbb9bd16a864c651109a0acbfc8aa6cef
Reviewed-on: https://go-review.googlesource.com/c/go/+/370703
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This makes testenv.RunWithTimeout first attempt to SIGQUIT the
subprocess to get a useful Go traceback, but if that doesn't work, it
sends a SIGKILL instead to make sure we tear down the subprocess. This
is potentially important for non-Go subprocesses.
For #37405.
Change-Id: I9e7e118dc5769ec3f45288a71658733bff30c9cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/370702
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This lifts the logic to run a subcommand with a timeout in a test from
the runtime's runTestProg into testenv. The implementation is
unchanged in this CL. We'll improve it in a future CL.
Currently, tests that run subcommands usually just timeout with no
useful output if the subcommand runs for too long. This is a step
toward improving this.
For #37405.
Change-Id: I2298770db516e216379c4c438e05d23cbbdda51d
Reviewed-on: https://go-review.googlesource.com/c/go/+/370701
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Weird, but don't drop the zone when stringifying.
Fixes#50111
Change-Id: I5fbccdfedcdc77a77ee6bafc8d82b8ec8ec7220c
Reviewed-on: https://go-review.googlesource.com/c/go/+/371094
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Matt Layher <mdlayher@gmail.com>
Trust: Matt Layher <mdlayher@gmail.com>
Trust: Ian Lance Taylor <iant@golang.org>
cmd/vet has several precision improvements for the checkers copylock, printf, sortslice, testinggoroutine, and tests. Adds a high level mention in the release notes and an example of string constant concatenation.
Updates #47694
Change-Id: I7a342a57ca3fd9e2f3e8ec99f7b647269798317f
Reviewed-on: https://go-review.googlesource.com/c/go/+/370734
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Trust: Tim King <taking@google.com>
Run-TryBot: Tim King <taking@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
The failure mode in #34264 appears to match
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=210726.
That bug was supposed to have been fixed in FreeBSD 12, but we're
still observing failures specifically for the 6-to-4 case on FreeBSD
12.2. It is not clear to me whether FreeBSD 13.0 is also affected.
For #34264
Change-Id: Iba7c7fc57676ae628b13c0b8fe43ddf2251c3637
Reviewed-on: https://go-review.googlesource.com/c/go/+/369157
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
Trust: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
When exporting the "any" empty interface type for unified IR, write it
out as a reference to the "any" alias, rather than to the underlying
empty interface. This matches how "byte" and "rune" are handled.
Verified to fix the issue demonstrated in CL 369975.
Change-Id: Ic2844b0acc3b17c20b3a40aaf262f62ec653eb5a
Reviewed-on: https://go-review.googlesource.com/c/go/+/370374
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
The 'Does not match' type error shoud be reported where
the function is called, not where the function is declared.
And fix the todo by gri of issue45985
Fixes#45985Fixes#49800
Change-Id: I15aac44dd44f2a57c485a1c273fcd79db912c389
Reviewed-on: https://go-review.googlesource.com/c/go/+/362634
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
The "-2008" builders are the only ones on which the failure has
been observed, so I suspect that it is due to a platform bug fixed in a
subsequent release.
Since no one has added a workaround since #19564 was filed over four
years ago, I'm assuming that no workaround is planned for this issue.
Let's add a skip for the known failure mode and call it at that.
Fixes#19564
Change-Id: Iefc22d1cc78bfdc79c845eb60cac22e26caf388c
Reviewed-on: https://go-review.googlesource.com/c/go/+/370377
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Generally speaking Go functions make no guarantees
about what has happened to result parameters on error,
and Pipe is no exception: callers should avoid looking at
p if Pipe returns an error.
However, we had a bug in which ForkExec was using the
content of p after a failed Pipe, and others may too.
As a robustness fix, make Pipe avoid writing to p on failure.
Updates #50057
Change-Id: Ie8955025dbd20702fabadc9bbe1d1a5ac0f36305
Reviewed-on: https://team-review.git.corp.google.com/c/golang/go-private/+/1291271
Reviewed-by: Ian Lance Taylor <iant@google.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/370577
Run-TryBot: Filippo Valsorda <filippo@golang.org>
Trust: Russ Cox <rsc@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Alex Rakoczy <alex@golang.org>
Pipe (and therefore forkLockPipe) does not make any guarantees
about the state of p after a failed Pipe(p). Avoid that assumption
and the too-clever goto, so that we don't accidentally Close a real fd
if the failed pipe leaves p[0] or p[1] set >= 0.
Fixes#50057
Fixes CVE-2021-44717
Change-Id: Iff8e19a6efbba0c73cc8b13ecfae381c87600bb4
Reviewed-on: https://team-review.git.corp.google.com/c/golang/go-private/+/1291270
Reviewed-by: Ian Lance Taylor <iant@google.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/370576
Run-TryBot: Filippo Valsorda <filippo@golang.org>
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Alex Rakoczy <alex@golang.org>
Also update some other go/types release notes to use the present tense.
Updates #47694
Change-Id: I654371c065e76fd5d22679d0d3c1a81bc3d1e513
Reviewed-on: https://go-review.googlesource.com/c/go/+/370235
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
TestLabelSystemstack needs to collect samples within runtime.systemstack
to complete the test.
The current approach uses fmt.Fprintf, which gets into systemstack
through the allocator and GC, but also does lots of other work. In my
measurements, approximately 2% of samples contain runtime.systemstack.
The new approach uses debug.SetGCPercent, which uses systemstack for
most of its work, including contention on mheap_.lock, which extends
usage even more. In my measurements, approximately 99% of samples
contain runtime.systemstack.
Fixes#50050
Change-Id: I59e5bb756341b716a12e13d2e3fe0adadd7fe956
Reviewed-on: https://go-review.googlesource.com/c/go/+/370375
Reviewed-by: Bryan Mills <bcmills@google.com>
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
We currently run one 'go list' invocation per GOMAXPROC. Since the go
command uses memory and has its own internal parallelism, that's
unlikely to be an efficient use of resources. Run half as many. I
suspect that's still too many but this should fix our OOMs.
For #49957.
Change-Id: Id06b6e0f0d96387a2a050e400f38bde6ba71aa60
Reviewed-on: https://go-review.googlesource.com/c/go/+/370376
Trust: Heschi Kreinick <heschi@google.com>
Run-TryBot: Heschi Kreinick <heschi@google.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Note when a statement mark was not consumed during Prog
generation, and try to use it on a subsequent opcode so
that the statement marker will not be lost.
And a test.
Fixes#49628.
Change-Id: I03f7782a9809cc4a0a5870df92b3e182cf124554
Reviewed-on: https://go-review.googlesource.com/c/go/+/366694
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Augmented some of the typeswitch*.go tests so that some instantiations
have duplicate cases, in order to ensure we're testing that.
Spacing changes in the tests are due to gofmt.
Change-Id: I5d3678813505c520c544281d4ac8a62ce7e236ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/370155
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
golang.org/cl/365234 incorrectly had pruningForGoVersion always return
workspace pruning instead of just returning workspace pruning at the top
level, which broke the proper determination of pruning for dependency
packages. Fix that code, and also fix a hang that resulted because the
module loading code keeps loading dependencies until it reaches a pruned
module or an unpruned module it already saw, so it could get stuck in a
cycle.
Change-Id: I8911f7d83aaee5870c43ef0355abbd439f15d4f7
Reviewed-on: https://go-review.googlesource.com/c/go/+/366775
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This is a clean port of CL 369434 to types2.
Change-Id: I3f9f80757bfbefb7b0417eef9e7b7c74c4c100b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/369474
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Our calculation of initOrder builds the dependency graph and then
removes function nodes approximately at random. While profiling, I
noticed that this latter step introduces a superlinear algorithm into
our type checking pass, which can dominate type checking for large
packages such as runtime.
It is hard to analyze this rigorously, but to give an idea of how such a
non-linearity could arise, suppose the following assumptions hold:
- Every function makes D calls at random to other functions in the
package, for some fixed constant D.
- The number of functions is proportional to N, the size of the package.
Under these simplified assumptions, the cost of removing an arbitrary
function F is P*D, where P is the expected number of functions calling
F. P has a Poisson distribution with mean D.
Now consider the fact that when removing a function F in position i, we
recursively pay the cost of copying F's predecessors and successors for
each node in the remaining unremoved subgraph of functions containing F.
With our assumptions, the size of this subgraph is proportional to
(N-i), the number of remaining functions to remove.
Therefore, the total cost of removing functions is proportional to
P*D*Σᴺ(N-i)
which is proportional to N².
However, if we remove functions in ascending order of cost, we can
partition by the number of predecessors, and the total cost of removing
functions is proportional to
N*D*Σ(PMF(X))
where PMF is the probability mass function of P. In other words cost is
proportional to N.
Assuming the above analysis is correct, it is still the case that the
initial assumptions are naive. Many large packages are more accurately
characterized as combinations of many smaller packages. Nevertheless, it
is intuitively clear that removing expensive nodes last should be
cheaper.
Therefore, we sort by cost first before removing nodes in
dependencyGraph.
We also move deletes to the outer loop, to avoid redundant deletes. By
inspection, this avoids a bug where n may not have been removed from its
successors if n had no predecessors.
name old time/op new time/op delta
Check/runtime/funcbodies/noinfo-8 568ms ±25% 82ms ± 1% -85.53% (p=0.000 n=8+10)
name old lines/s new lines/s delta
Check/runtime/funcbodies/noinfo-8 93.1k ±56% 705.1k ± 1% +657.63% (p=0.000 n=10+10)
Updates #49856
Change-Id: Id2e70d67401af19205e1e0b9947baa16dd6506f0
Reviewed-on: https://go-review.googlesource.com/c/go/+/369434
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Preparation for #44505, but safe for Go 1.18.
Also fixes the default build on Macs, at least for
people who have a $HOME/go1.17 or have run
go install golang.org/dl/go1.17@latest
go1.17 download
Replay of CL 369914 after revert in CL 370138.
Only change is adding 'export GOROOT_BOOTSTRAP' in make.bash.
Change-Id: I8ced4e87a9dc0f05cc49095578b401ae6212ac85
Reviewed-on: https://go-review.googlesource.com/c/go/+/370274
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
If the test's main goroutine receives a SIGPROF while creating the
C-owned thread for the test, that sample will appear in the resulting
profile. The root end of that stack will show a set of Go functions. The
leaf end will be the C functions returned by the SetCgoTraceback
handler, which will confuse the test runner.
Add a label to the main goroutine while it calls in to C, so all profile
samples that triggered the SetCgoTraceback handler are either correct,
or can easily be excluded from the test's analysis. (The labels will not
apply to the resulting C-owned thread, which does not use goroutines.)
Fixes#43174
Change-Id: Ica3100ca0f191dcf91b30b0084e8541c5a25689f
Reviewed-on: https://go-review.googlesource.com/c/go/+/370135
Reviewed-by: Michael Pratt <mpratt@google.com>
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This test seems like it needs attention from a TLS and/or FreeBSD
expert. In the meantime, it needs to stop causing noise on the build
dashboard.
For #19229
Change-Id: If7e9e3533ae7cb29006a670c3e9df90512dcf9f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/370137
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This reverts CL 366536
Reason for revert: may have caused #50033 due to an invalid or partially-populated *TCPAddr
Fixes#50033
Change-Id: Ia29ca4116503dba65d56e89caa46ba1c848d421a
Reviewed-on: https://go-review.googlesource.com/c/go/+/369982
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
As a side effect of the changes in cmd/go/internal/work in CL 369977,
binaries built in GOPATH mode now include rudimentary build metadata
for at least the package path and compiler in use.
That seems like a strict improvement, but the test needs to be updated
to reflect the newly-available metadata.
Change-Id: I657c785e3e9992ed594c9524409f2d076f9eb376
Reviewed-on: https://go-review.googlesource.com/c/go/+/370234
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
