Change-Id: I3bcf8850ad3873f2627ba017cbfb8b7a8c9cf467
Reviewed-on: https://go-review.googlesource.com/125256
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
If we have go get -u x1@v1 x2@v2 and x1 depends on x2,
use v2 as the "upgraded" x2 chosen by -u instead of
letting -u pick something (say, v2.1) and then immediately
overriding it. This avoids chasing down the deps from v2.1
and also avoids them polluting the overall module graph.
This fix also lets us delete some code in the preparation step,
reducing the overall latency of get -u.
Suggested by Bryan Mills in
https://go-review.googlesource.com/c/vgo/+/122396/6#371.
Fixes#26342.
Change-Id: I50fa842304820d3f16f66a8e65dea695e2b0f88b
Reviewed-on: https://go-review.googlesource.com/124856
Reviewed-by: Bryan C. Mills <bcmills@google.com>
If x v1.0.0 requires y v1.2.0, then
go get x@v1.0.0 y@v1.0.0 needs to fail gracefully.
Fixes#25917.
Change-Id: I9b426af23a30310fcb0c3545a8d97feb58b8ddbe
Reviewed-on: https://go-review.googlesource.com/124800
Reviewed-by: Bryan C. Mills <bcmills@google.com>
While writing the GOPROXY docs it occurred to me that versions
can contain upper-case letters as well. The docs therefore say
that versions are case-encoded the same as paths in the proxy
protocol (and therefore in the cache as well). Make it so.
Change-Id: Ibc0c4af0192a4af251e5dd6f2d36cda7e529099a
Reviewed-on: https://go-review.googlesource.com/124795
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Use the dedicated AES* and PMULL* instructions to accelerate AES-GCM
name old time/op new time/op delta
AESGCMSeal1K-46 12.1µs ± 0% 0.9µs ± 0% -92.66% (p=0.000 n=9+10)
AESGCMOpen1K-46 12.1µs ± 0% 0.9µs ± 0% -92.43% (p=0.000 n=10+10)
AESGCMSign8K-46 58.6µs ± 0% 2.1µs ± 0% -96.41% (p=0.000 n=9+8)
AESGCMSeal8K-46 92.8µs ± 0% 5.7µs ± 0% -93.86% (p=0.000 n=9+9)
AESGCMOpen8K-46 92.9µs ± 0% 5.7µs ± 0% -93.84% (p=0.000 n=8+9)
name old speed new speed delta
AESGCMSeal1K-46 84.7MB/s ± 0% 1153.4MB/s ± 0% +1262.21% (p=0.000 n=9+10)
AESGCMOpen1K-46 84.4MB/s ± 0% 1115.2MB/s ± 0% +1220.53% (p=0.000 n=10+10)
AESGCMSign8K-46 140MB/s ± 0% 3894MB/s ± 0% +2687.50% (p=0.000 n=9+10)
AESGCMSeal8K-46 88.2MB/s ± 0% 1437.5MB/s ± 0% +1529.30% (p=0.000 n=9+9)
AESGCMOpen8K-46 88.2MB/s ± 0% 1430.5MB/s ± 0% +1522.01% (p=0.000 n=8+9)
This change mirrors the current amd64 implementation, and provides optimal performance
on a range of arm64 processors including Centriq 2400 and Apple A12. By and large it is
implicitly tested by the robustness of the already existing amd64 implementation.
The implementation interleaves GHASH with CTR mode to achieve the highest possible
throughput, it also aggregates GHASH with a factor of 8, to decrease the cost of the
reduction step.
Even thought there is a significant amount of assembly, the code reuses the go
code for the amd64 implementation, so there is little additional go code.
Since AES-GCM is critical for performance of all web servers, this change is
required to level the playfield for arm64 CPUs, where amd64 currently enjoys an
unfair advantage.
Ideally both amd64 and arm64 codepaths could be replaced by hypothetical AES and
CLMUL intrinsics, with a few additional vector instructions.
Fixes#18498Fixes#19840
Change-Id: Icc57b868cd1f67ac695c1ac163a8e215f74c7910
Reviewed-on: https://go-review.googlesource.com/107298
Run-TryBot: Vlad Krasnov <vlad@cloudflare.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
I forgot to run trybots on CL 123758, and the test failed on Windows because I
hard-coded a slash-delimited path.
Use the tent-in-a-box operator ([/\\]) to make the path platform-agnostic.
Change-Id: I9113ab60d21152c11e2ebdf822b58a44b1b38574
Reviewed-on: https://go-review.googlesource.com/125115
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The test runs far too long for -short mode (4 seconds).
Also removed useless test of now-disconnected knob
(GO_SSA_PHI_LOC_CUTOFF), which cuts 4 seconds to 2 seconds (which
is still too long), and finished removing the disconnected knob.
Updates #26469.
Change-Id: I6c594227c4a5aaffee46832049bdbbf570d86e60
Reviewed-on: https://go-review.googlesource.com/125075
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
If the the package cannot be built,
"go test" and "go vet" should not run the "vet" tool.
In that case only errors from the compilers will be displayed.
Fixes#26125
Change-Id: I5da6ba64bae5f44feaf5bd4e765eea85533cddd4
Reviewed-on: https://go-review.googlesource.com/123938
Run-TryBot: Yury Smolsky <yury@smolsky.by>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
CL 123577 added TestScript. The install_rebuild_gopath
test was failing on Plan 9 because it defines a GOPATH
using the ':' separator, while Plan 9 expects the '\000'
separator in environment variables.
This change fixes the script engine by defining a new
":" environment variable set to OS-specific path list
separator.
The install_rebuild_gopath test has been updated to use
"${:}" instead of ":".
Fixes#26421.
Change-Id: I58a97f882cdb48cc0836398b0d98a80ea58041ba
Reviewed-on: https://go-review.googlesource.com/124435
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
When using callbacks, it is not necessarily a deadlock if there is no
runnable goroutine, since a callback might still be pending. If there
is no callback pending, Node.js simply exits with exit code zero,
which is not desired if the Go program is still considered running.
This is why an explicit check on exit is used to trigger the "deadlock"
error. This CL makes it so this is Go's normal "deadlock" error, which
includes the stack traces of all goroutines.
Updates #26382
Change-Id: If88486684d0517a64f570009a5ea0ad082679a54
Reviewed-on: https://go-review.googlesource.com/123936
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
It's important for a smooth transition for non-module users
not to change operation in GOPATH/src by default in Go 1.11,
even if go.mod exists in a downloaded dependency.
Even so, users create go.mod and then are confused about
why 'go get' commands seem to behave oddly, when in fact
they are getting the old 'go get'.
Try to split the difference by printing a warning in 'go get'
when run in a tree that would normally be considered a
module if only it were outside GOPATH/src.
Change-Id: I55a1cbef127f3f36de54a8d7b93e1fc64bf0a708
Reviewed-on: https://go-review.googlesource.com/124859
Reviewed-by: Bryan C. Mills <bcmills@google.com>
If there was a mod and non-mod meta tag for a given prefix,
the meta tag extractor was already dropping the non-mod meta tag.
But we might have mod and non-mod meta tags with different
prefixes, in which case the mod tag should prevail when both match.
Fixes#26200.
Change-Id: I17ab361338e270b9fa03999ad1954f2bbe0f5017
Reviewed-on: https://go-review.googlesource.com/124714
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Updates http2 to x/net/http2 git rev a680a1efc54 for:
http2: reject large SETTINGS frames or those with duplicates
https://golang.org/cl/124735
Change-Id: I2168d1d1eef9c63b1a9c06b514b77fae16f920ed
Reviewed-on: https://go-review.googlesource.com/125036
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This was an unfortunate debugging print introduced
while working on the unfortunately large CL 123576.
At least now we're done with that awfulness.
Change-Id: Ib83db59382a799f649832d22d3c6f039d2ef9d2c
Reviewed-on: https://go-review.googlesource.com/125015
Reviewed-by: Bryan C. Mills <bcmills@google.com>
go.sum accumulates cruft as modules are added and removed as
direct and indirect dependencies. Instead of exposing all that cruft,
let "go mod -sync" clean it out.
Fixes#26381.
Change-Id: I7c9534cf7cc4579f7f82646d00ff691c87a13c4a
Reviewed-on: https://go-review.googlesource.com/124713
Reviewed-by: Bryan C. Mills <bcmills@google.com>
People are (understandably) confused by creating go.mod files in GOPATH/src
and then having the go command not use modules in those directories.
We can't change that behavior (or we'll break non-module users of GOPATH)
but we can force 'go mod' (including 'go mod -init') to fail loudly in that case.
If this is not enough, the next step would be to print a warning every time
the go command is run in a GOPATH/src directory with a go.mod but
module mode hasn't triggered. But that will annoy all the non-module users.
Hopefully anyone confused will eventually run a 'go mod' command of
some kind, which will fail loudly.
Fixes#26365.
Change-Id: I8c5fe987fbc3f8d2eceb1138e6862a391ade150c
Reviewed-on: https://go-review.googlesource.com/124708
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Also document module use of GOPATH including GOPATH/src/mod
and GOPATH/bin (unless GOBIN is set).
Fixes#26399.
Fixes#26406.
Change-Id: I7be8eaf110f4fa6fc76ea4cd39aea3dd8addf0b0
Reviewed-on: https://go-review.googlesource.com/124707
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Now pkg@none actually removes the pkg instead of dying.
For #26342.
Change-Id: I9df7281ed8fd24480109b36f33a563f92e279244
Reviewed-on: https://go-review.googlesource.com/124796
Reviewed-by: Bryan C. Mills <bcmills@google.com>
The original pseudo-version design used versions of the form
v0.0.0-yyyymmddhhmmss-abcdef123456
These were intentionally chosen to be valid semantic versions
that sort below any explicitly-chosen semantic version (even v0.0.0),
so that they could be used before anything was tagged but after
that would essentially only be useful in replace statements
(because the max operation during MVS would always prefer
a tagged version).
Then we changed the go command to accept hashes on the
command line, so that you can say
go get github.com/my/proj@abcdef
and it will download and use v0.0.0-yyyymmddhhmmss-abcdef123456.
If you were using v1.10.1 before and this commit is just little bit
newer than that commit, calling it v0.0.0-xxx is confusing but
also harmful: the go command sees the change from v1.10.1 to
the v0.0.0 pseudoversion as a downgrade, and it downgrades other
modules in the build. In particular if some other module has
a requirement of github.com/my/proj v1.9.0 (or later), the
pseudo-version appears to be before that, so go get would
downgrade that module too. It might even remove it entirely,
if every available version needs a post-v0.0.0 version of my/proj.
This CL introduces new pseudo-version forms that can be used
to slot in after the most recent explicit tag before the commit.
If the most recent tagged commit before abcdef is v1.10.1,
then now we will use
v1.10.2-0.yyyymmddhhmmss-abcdef123456
This has the right properties for downgrades and the like,
since it is after v1.10.1 but before almost any possible
successor, such as v1.10.2, v1.10.2-1, or v1.10.2-pre.
This CL also uses those pseudo-version forms as appropriate
when mapping a hash to a pseudo-version. This fixes the
downgrade problem.
Overall, this CL reflects our growing recognition of pseudo-versions
as being like "untagged prereleases".
Issue #26150 was about documenting best practices for how
to work around this kind of accidental downgrade problem
with additional steps. Now there are no additional steps:
the problem is avoided by default.
Fixes#26150.
Change-Id: I402feeccb93e8e937bafcaa26402d88572e9b14c
Reviewed-on: https://go-review.googlesource.com/124515
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Repos written before the introduction of semantic import versioning
introduced tags like v2.0.0, v3.0.0, and so on, expecting that
(1) the import path would remain unchanged, and perhaps also
(2) there would be at most one copy of the package in a build.
We've always accommodated these by mapping them into the
v0/v1 version range, so that if you ran
go get k8s.io/client-go@v8.0.0
it would not complain about v8.x.x being a non-v1 version and
instead would map that version to a pseudo-version in go.mod:
require k8s.io/client-go v0.0.0-20180628043050-7d04d0e2a0a1
The pseudo-version fails to capture two important facts: first,
that this really is the v8.0.0 tag, and second, that it should be
preferred over any earlier v1 tags.
A related problem is that running "go get k8s.io/client-go"
with no version will choose the latest v1 tag (v1.5.1), which
is obsolete.
This CL introduces a new version suffix +incompatible that
indicates that the tag should be considered an (incompatible)
extension of the v1 version sequence instead of part of its
own major version with its own versioned module path.
The requirement above can now be written:
require k8s.io/client-go v8.0.0+incompatible
(The +metadata suffix is a standard part of semantic versioning,
and that suffix is ignored when comparing two versions for
precedence or equality. As part of canonicalizing versions
recorded in go.mod, the go command has always stripped all
such suffixes. It still strips nearly all: only +incompatible is
preserved now.)
In addition to recognizing the +incompatible, the code that
maps a commit hash to a version will use that form when
appropriate, so that
go get k8s.io/client-go@7d04d0
will choose k8s.io/client-go@v8.0.0+incompatible.
Also, the code that computes the list of available versions from
a given source code repository also maps old tags to +incompatible
versions, for any tagged commit in which a go.mod file does not exist.
Therefore
go list -m -versions k8s.io/client-go@latest
will show
k8s.io/client-go v1.4.0 v1.5.0 v1.5.1 v2.0.0-alpha.0+incompatible ... v8.0.0+incompatible
and similarly
go get k8s.io/client-go
will now choose v8.0.0+incompatible as the meaning of "latest tagged version".
The extraction of +incompatible versions from source code repos
depends on a codehost.Repo method ReadFileRevs, to do a bulk read
of multiple revisions of a file. That method is only implemented for git in this CL.
Future CLs will need to add support for that method to the other repository
implementations.
Documentation for this change is in CL 124515.
Fixes#26238.
Change-Id: I5bb1d7a46b5fffde34a3c0e6f8d19d9608188cea
Reviewed-on: https://go-review.googlesource.com/124384
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Windows networking doesn't work without this environment variable (#25210).
Re-enable TestScript on Windows, and fix two minor failures.
Fixes#26457.
Change-Id: Id9bea49dfb58403195c29c3d831a532ef0f9a233
Reviewed-on: https://go-review.googlesource.com/124858
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TSAN for Go only supports heap address in the range [0x00c000000000,
0x00e000000000). However, we currently create heap hints of the form
0xXXc000000000 for XX between 0x00 and 0x7f. Even for XX=0x01, this
hint is outside TSAN's supported heap address range.
Fix this by creating a slightly different set of hints in race mode,
all of which fall inside TSAN's heap address range.
This should fix TestArenaCollision flakes. That test forces the
runtime to use later heap hints. Currently, this always results in
TSAN "failed to allocate" failures on Windows (which happens to have a
slightly more constrained TSAN layout than non-Windows). Most of the
time we don't notice these failures, but sometimes it crashes TSAN,
leading to a test failure.
Fixes#25698.
Change-Id: I8926cd61f0ee5ee00efa77b283f7b809c555be46
Reviewed-on: https://go-review.googlesource.com/123780
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
Code fix was in CL 122556. This is a corresponding test case.
Fixes#26426
Change-Id: Ib8769f367aed8bead029da0a8d2ddccee1d1dccb
Reviewed-on: https://go-review.googlesource.com/124535
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The backwards incompatible changes were undone in CL 120355, while still
preserving the additions needed for assignments in templates to work.
Change-Id: Ie76a798916ef36509c88e171a04bb2cf2a3d7e8e
Reviewed-on: https://go-review.googlesource.com/124917
Reviewed-by: Andrew Bonventre <andybons@golang.org>
Until we figure out how to deal with gdb on Darwin (doesn't
read compressed DWARF from binaries), avoid compressing
DWARF in that case so that the test will still yield meaningful
results.
This is also reported to be a problem for Windows.
Problem also exists for lldb, but this test doesn't check
lldb.
Updates #25925
Change-Id: I85c0e5db75f3329957290500626a3ac7f078f608
Reviewed-on: https://go-review.googlesource.com/124712
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Pointed out in CL 122396.
An empty prefix has already been handled above.
Change-Id: Ib94df0a9c8c0517f932b90126232111caa9ad289
Reviewed-on: https://go-review.googlesource.com/124797
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Completely replace the opener, which had become not only stale
but bad, expand the discussion of the gopher, and generally provide
prose more connected to the present than to the programming world
of 2007.
Fixes#26107
Change-Id: I5e72f0c81e71d1237fe142dc26114991329a6996
Reviewed-on: https://go-review.googlesource.com/124616
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The final API uses 'region' instead of 'span' from the proposal.
Change-Id: I305da891a360596fff89b10bc6de3090289b5396
Reviewed-on: https://go-review.googlesource.com/124815
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
If one tries to use promoted fields in a struct literal, the compiler
errors correctly. However, if the embedded fields are of struct pointer
type, the field.Type.Sym.Name expression below panics.
This is because field.Type.Sym is nil in that case. We can simply use
field.Sym.Name in this piece of code though, as it only concerns
embedded fields, in which case what we are after is the field name.
Added a test mirroring fixedbugs/issue23609.go, but with pointer types.
Fixes#26416.
Change-Id: Ia46ce62995c9e1653f315accb99d592aff2f285e
Reviewed-on: https://go-review.googlesource.com/124395
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Text based on CL 124655.
Change-Id: I7c4866ce829cb28a4c60cd8ced3ef99047a38c54
Reviewed-on: https://go-review.googlesource.com/124711
Reviewed-by: Austin Clements <austin@google.com>
If a filepath.WalkFunc is called with an non-nil err argument, it's possible
that the info argument will be nil. The comment above filepath.WalkFunc now
reflects this.
Fixes#26425
Change-Id: Ib9963b3344587d2993f1698c5a801f2d1286856b
GitHub-Last-Rev: 553fc266b5
GitHub-Pull-Request: golang/go#26435
Reviewed-on: https://go-review.googlesource.com/124635
Reviewed-by: Ian Lance Taylor <iant@golang.org>
