]> Cypherpunks repositories - gostls13.git/log
gostls13.git
4 years agocmd/compile: do not declare func nodes
Cuong Manh Le [Wed, 19 Aug 2020 09:05:02 +0000 (16:05 +0700)]
cmd/compile: do not declare func nodes

The primary responsibility of declare() to associate a symbol (Sym) with
a declaration (Node), so "oldname" will work. Function literals are
anonymous, so their symbols does not need to be declared.

Passes toolstash-check.

Change-Id: I739b1054e3953e85fbd74a99148b9cfd7e5a57eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/249078
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/compile: do not push functions literal node to Func.Dcl
Cuong Manh Le [Wed, 2 Sep 2020 05:45:07 +0000 (12:45 +0700)]
cmd/compile: do not push functions literal node to Func.Dcl

They are skipped while processing Func.Dcl anyway.

This CL does not pass toolstash-check, because it reduces the length
of Func.Dcl length, while that length is used to generate autotmp
variables name.

Change-Id: I408183e62ce6c34e5f04c89814ebb9570957e37b
Reviewed-on: https://go-review.googlesource.com/c/go/+/252418
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/go: add -Wl,-Bsymbolic-functions to cgo flags whitelist
Dan Kortschak [Wed, 2 Sep 2020 22:57:34 +0000 (08:27 +0930)]
cmd/go: add -Wl,-Bsymbolic-functions to cgo flags whitelist

Closes #41199

Change-Id: Iab69358e8c39e6d2b2797c7ce750df63aa7e96b0
Reviewed-on: https://go-review.googlesource.com/c/go/+/252698
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agonet/http: reject negative suffix-length Range:bytes=--N with 416 status code
Emmanuel T Odeke [Wed, 2 Sep 2020 08:08:02 +0000 (01:08 -0700)]
net/http: reject negative suffix-length Range:bytes=--N with 416 status code

Fixes the file server to reject requests of the form:
    "Range": "bytes=--N"
where "-N" is a negative suffix-length as designated by the
grammar in RFC 7233 Section 2.1, "Byte-Ranges", which specifies
that suffix-length MUST be of the form 1*DIGIT aka a non-negative digit.

Thus requests such as:
    "Range": "bytes=--2"
will be rejected with a "416 Range Not Satisfiable" response.

Fixes #40940

Change-Id: I3e89f8326c14af30d8bdb126998a50e02ba002d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/252497
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile/internal/gc: remove unparen
Cuong Manh Le [Wed, 2 Sep 2020 15:10:15 +0000 (22:10 +0700)]
cmd/compile/internal/gc: remove unparen

CL 197120 removed the last use of it.

Change-Id: I5fe4f57a47acc712208d831e72cd79205a534c28
Reviewed-on: https://go-review.googlesource.com/c/go/+/252697
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agotesting: flush test summaries to stdout atomically when streaming output
Bryan C. Mills [Wed, 19 Aug 2020 02:47:12 +0000 (22:47 -0400)]
testing: flush test summaries to stdout atomically when streaming output

While debugging #40771, I realized that the chatty printer should only
ever print to a single io.Writer (normally os.Stdout). The other
Writer implementations in the chain write to local buffers, but if we
wrote a test's output to a local buffer, then we did *not* write it to
stdout and we should not store it as the most recently logged test.

Because the chatty printer should only ever print to one place, it
shouldn't receive an io.Writer as an argument — rather, it shouldn't
be used at all for destinations other than the main output stream.

On the other hand, when we flush the output buffer to stdout in the
top-level flushToParent call, it is important that we not allow some
other test's output to intrude between the test summary header and the
remainder of the test's output. cmd/test2json doesn't know how to
parse such an intrusion, and it's confusing to humans too.

No test because I couldn't reproduce the user-reported error without
modifying the testing package. (This behavior seems to be very
sensitive to output size and/or goroutine scheduling.)

Fixes #40771
Updates #38458

Change-Id: Ic19bf1d535672b096ba1c8583a3b74aab6d6d766
Reviewed-on: https://go-review.googlesource.com/c/go/+/249026
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoruntime: add file copyright header declaration
Heisenberg [Mon, 3 Aug 2020 07:33:47 +0000 (15:33 +0800)]
runtime: add file copyright header declaration

Some files have no copyright notice.
The copyright time is the earliest modification record of the file.

Change-Id: I5698bae16b6b73543e074415877a03348f792951
Reviewed-on: https://go-review.googlesource.com/c/go/+/246378
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
4 years agonet/http/fcgi: fix race in child.serve connection read
Tzu-Chiao Yeh [Wed, 2 Sep 2020 02:53:39 +0000 (10:53 +0800)]
net/http/fcgi: fix race in child.serve connection read

Guards the connection read with a mutex, because typeStdin
asynchronously and concurrently writes to the underlying conn.

Fixes #41167

Change-Id: Ia2610f4fde0bd4b108c54164095ea293980b0301
Reviewed-on: https://go-review.googlesource.com/c/go/+/252417
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
4 years agoruntime: fix panic if newstack at runtime.acquireLockRank
chainhelen [Fri, 21 Aug 2020 16:44:52 +0000 (16:44 +0000)]
runtime: fix panic if newstack at runtime.acquireLockRank

Process may crash becaues acquireLockRank and releaseLockRank may
be called in nosplit context. With optimizations and inlining
disabled, these functions won't get inlined or have their morestack
calls eliminated.
Nosplit is not strictly required for lockWithRank, unlockWithRank
and lockWithRankMayAcquire, just keep consistency with lockrank_on.go
here.

Fixes #40843

Change-Id: I5824119f98a1da66d767cdb9a60dffe768f13c81
GitHub-Last-Rev: 38fd3ccf6ea03b670c7561c060ccdbccc42fff40
GitHub-Pull-Request: golang/go#40844
Reviewed-on: https://go-review.googlesource.com/c/go/+/248878
Reviewed-by: Dan Scales <danscales@google.com>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocmd/cgo: document #include <> search path behaviour
KJ Tsanaktsidis [Tue, 1 Sep 2020 00:00:23 +0000 (10:00 +1000)]
cmd/cgo: document #include <> search path behaviour

cgo effectively prepends -I${SRCDIR} to the header include path of all
preambles it processes, so when an #include <> matches a header file
both in the source directory and also another include directory, the
local copy will be used in preference.

This behaviour is surprising but unfortunately also longstanding and
relied upon by packages in the wild, so the best we can do is to
document it.

Fixes #41059

Change-Id: If6d2818294b2bd94ea0fe5fd6ce77e54b3e167a6
Reviewed-on: https://go-review.googlesource.com/c/go/+/251758
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agonet/smtp: adds support for the SMTPUTF8 extension
Daniel Cormier [Tue, 1 Sep 2020 21:13:01 +0000 (21:13 +0000)]
net/smtp: adds support for the SMTPUTF8 extension

If the SMTP server supports the SMTPUTF8 extension,
the SMTPUTF8 parameter is added to the MAIL FROM
command by the (*Client).Mail method.

Fixes #19860

Change-Id: I3287faf114ee514e5faa815a6bbc1bf04cf60b0f
GitHub-Last-Rev: d6338bb802da7537223f1ec6eda960606febefb8
GitHub-Pull-Request: golang/go#40627
Reviewed-on: https://go-review.googlesource.com/c/go/+/247257
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/go: simplify and document lazy-loading test cases
Bryan C. Mills [Tue, 1 Sep 2020 03:15:05 +0000 (23:15 -0400)]
cmd/go: simplify and document lazy-loading test cases

I decided to add package and module diagrams to the test cases to make
them easier to follow.

While adding those diagrams, I noticed some strong similarities among
a couple of the graphs, so I consolidated those cases (and deleted the
redundant tests).

For #36460

Change-Id: Id6cd04fc871379b83851c2d1af89ea9296a0f3e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/251997
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agonet/http/cgi: don't pass nil Body to the child handler
Marco [Sun, 24 May 2020 12:58:51 +0000 (14:58 +0200)]
net/http/cgi: don't pass nil Body to the child handler

For server requests, the http.Request Body should not be nil.

Fixes #39190

Change-Id: I32de7b6c0f6ca55008fea9fd86089cda0a2dea62
Reviewed-on: https://go-review.googlesource.com/c/go/+/235137
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocmd/go: remove TestScript/get_unicode and an internal get test case
Michael Matloob [Tue, 1 Sep 2020 19:41:36 +0000 (15:41 -0400)]
cmd/go: remove TestScript/get_unicode and an internal get test case

That test tested that import paths with non-ASCII unicode paths
were allowed by the Go command. Remove this test case because
golang.org/cl/251878 removes that support.

Also rewrite a test case in TestRepoRootForImportPath in the test
for cmd/go/internal/get to reflect that unicode directory names are now
disallowed.

Updates #29101

Change-Id: I669e220facd04fc82ccd05dd08e8f1ff4d48b1fd
Reviewed-on: https://go-review.googlesource.com/c/go/+/252297
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agotesting: fail Example tests that invoke runtime.Goexit
Changkun Ou [Mon, 31 Aug 2020 18:54:17 +0000 (20:54 +0200)]
testing: fail Example tests that invoke runtime.Goexit

Previously, if an example test invoked runtime.Goexit, it would
pass yet hang until a timeout, while regular tests that invoke
runtime.Goexit do fail. This change removes that inconsistent
behavior and makes such example tests fail, and panic with an
indication of having invoked runtime.Goexit.

Fixes #41084

Change-Id: I0ffa152204f2b1580f4d5d6961ba1ce6b13fc022
Reviewed-on: https://go-review.googlesource.com/c/go/+/251857
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocmd/go/internal/get: add GOINSECURE support
witchard [Sun, 30 Aug 2020 18:15:03 +0000 (18:15 +0000)]
cmd/go/internal/get: add GOINSECURE support

Adds support for the GOINSECURE environment variable to GOPATH mode.

Updates #37519.

Change-Id: Ibe3f52b7f30b1395edb000998905ee93abe6cada
GitHub-Last-Rev: e298c0009eb5eba537bb00185a8778d2aab696ba
GitHub-Pull-Request: golang/go#38628
Reviewed-on: https://go-review.googlesource.com/c/go/+/229758
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go/internal/get: disallow non-ASCII unicode letters from import paths
Michael Matloob [Mon, 31 Aug 2020 20:12:35 +0000 (16:12 -0400)]
cmd/go/internal/get: disallow non-ASCII unicode letters from import paths

The copy of CheckImportPath in path.go and the regular expression for github
repos in vcsPaths together allow import paths with unicode letters with import
paths. These all come from github repos with non-ASCII unicode letters
with paths in directories. This mainly shows up in GOPATH mode, but could
also show up in Module mode when getting a module in GOPROXY=direct mode.

We expect there to not be any significant affected users of this change--
an investingation of github repos that would produce import paths that
would comply with the copy CheckImportPaths that's being removed, but not
modload.CheckImportPaths only surfaced a handful of cases, all of which
seemed to be small test or demonstation repos. But this CL is being
submitted early in the cycle so that it can be backed out if need be.

Updates #29101

Change-Id: I719df4af5b318e1330e90d8a0bffe5bb8d816f4f
Reviewed-on: https://go-review.googlesource.com/c/go/+/251878
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd: update vendored golang.org/x/mod
Michael Matloob [Mon, 31 Aug 2020 20:29:13 +0000 (16:29 -0400)]
cmd: update vendored golang.org/x/mod

This pulls in golang.org/cl/250920 which rejects Windows shortnames as
path components in module.CheckImportPath (as is already done in
cmd/go/internal/get's copy of CheckImportPath). This will allow us to replace
the copy of CheckImportPath with the original.

This also pulls in golang.org/cl/250919 which rejects + in CheckPath and
CheckImportPath, and golang.org/cl/235597, which adds methods to the zip
package for gorelease, but shouldn't affect cmd.

This change also updates the cmd/go test case TestScript/mod_bad_filenames
to reflect that golang.org/x/mod/zip error messages now include filenames
for bad file names that can't be included in zip archives.

Updates #29101

Change-Id: I7f654325dc33b19bc9c6f77a56546747add5a47f
Reviewed-on: https://go-review.googlesource.com/c/go/+/251877
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agodoc/asm: add BP is callee-save paragraph
Egon Elbre [Tue, 25 Aug 2020 07:38:50 +0000 (10:38 +0300)]
doc/asm: add BP is callee-save paragraph

Change-Id: Id38e639c66a42acf0b1c4488cdfd0b7b6cf71c78
Reviewed-on: https://go-review.googlesource.com/c/go/+/250397
Reviewed-by: Keith Randall <khr@golang.org>
4 years agoruntime: remove remnants of signal stack workaround
Heisenberg [Mon, 31 Aug 2020 11:22:36 +0000 (19:22 +0800)]
runtime: remove remnants of signal stack workaround

Updates #35979

Change-Id: Ic3a6e1b5e9d544979a3c8d909a36a55efa3b9c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/251757
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
4 years agonet/http: refactor ResponseWriter.ReadFrom to permit splice on Linux
Paul Forgey [Tue, 1 Sep 2020 00:38:01 +0000 (00:38 +0000)]
net/http: refactor ResponseWriter.ReadFrom to permit splice on Linux

Rather than probe and guess if sendfile will work inside ResponseWriter.ReadFrom(src),
this change fixes the underlying issue of starting to respond before src is readable
We'll no longer send a status OK if a header has not yet been written and reading
from src is destined to fail. This small change implicitly takes care of the need for
the server to sniff the response body to determine the Content-Type.

This allows splice to work on Linux when src is a socket or any non-regular file that's spliceable.

The extra read of 512 bytes may raise an objection, and that's fair, but
we're already swapping some syscall prep work for another and a read of
512 probably will not impact the overall performance. For shorter
bodies, there's likely less setup time. A little initial slop is not too
unusual in zero copy network code, and sometimes actually helps.

Fixes #40888

Change-Id: I4a8e2ad0ace1318bae66dae5671d06ea6d4838ed
GitHub-Last-Rev: 097364ea866613d103a31e2247b44f4a12077f9e
GitHub-Pull-Request: golang/go#40903
Reviewed-on: https://go-review.googlesource.com/c/go/+/249238
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
4 years agodebug/elf: support relocations relative to sections with non-zero addresses
Victor Michel [Sun, 30 Aug 2020 20:43:39 +0000 (20:43 +0000)]
debug/elf: support relocations relative to sections with non-zero addresses

commit 72ec930fa70c20ce69b21bf32a7916c04c2e9c2f added basic support for
relocations, but assumed that the symbol value would be 0, likely because
.debug_info always has address == 0 in the ELF section headers.

CL 195679 added further support for relocations, but explicitly encoded
the original assumption that section addresses would be 0.

This change removes that assumption: all relocations will now be
properly computed based on the target symbol value even when that symbol
is a section with a non-zero address.

Typically, sections that are part of a LOAD program segment have
non-zero addresses. For example, .debug_ranges relocations could be
relative to .text, which usually has an address > 0.

Fixes #40879

Change-Id: Ib0a616bb8b05d6c96d179b03ca33a10946fc5d59
GitHub-Last-Rev: 4200de732641995f3a4958a13a5c78f65b7eae50
GitHub-Pull-Request: golang/go#41038
Reviewed-on: https://go-review.googlesource.com/c/go/+/250559
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/compile: adjust some AMD64 rewrite rules to use typed aux fields
David Finkel [Sat, 25 Apr 2020 17:32:06 +0000 (13:32 -0400)]
cmd/compile: adjust some AMD64 rewrite rules to use typed aux fields

Remove an extra int32-representable check when deciding to use an int32
constant as an immediate value.

Comment out a broken optimization that relies on MaxUint32 being
representable by a signed int32. It never triggers and when fixed, the
signedness of the auxint prevents other optimization passes from
handling it properly, thus causing segfaults in the runtime.

Remove a couple offset representable in 32-bits checks on 32-bit aux
vals.

toolstash-check clean

Change-Id: I148b53403fde523c90d692cb90e412460664b439
Reviewed-on: https://go-review.googlesource.com/c/go/+/230458
Reviewed-by: Keith Randall <khr@golang.org>
4 years agocmd/compile,cmd/asm: simplify recording of branch targets, take 2
Keith Randall [Fri, 28 Aug 2020 17:10:32 +0000 (17:10 +0000)]
cmd/compile,cmd/asm: simplify recording of branch targets, take 2

We currently use two fields to store the targets of branches.
Some phases use p.To.Val, some use p.Pcond. Rewrite so that
every branch instruction uses p.To.Val.
p.From.Val is also used in rare instances.
Introduce a Pool link for use by arm/arm64, instead of
repurposing Pcond.

This is a cleanup CL in preparation for some stack frame CLs.

Change-Id: If8239177e4b1ea2bccd0608eb39553d23210d405
Reviewed-on: https://go-review.googlesource.com/c/go/+/251437
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agodebug/elf: run relocation tests in parallel
Ian Lance Taylor [Sat, 29 Aug 2020 21:11:18 +0000 (14:11 -0700)]
debug/elf: run relocation tests in parallel

Also don't restart DWARF reading from beginning when we are testing
multiple entries.

Also reformat relocationTests slice to use indexed literals.

Change-Id: Ia5f17214483394d0ef033be516df61f0bdc521b6
Reviewed-on: https://go-review.googlesource.com/c/go/+/251637
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
4 years agodatabase/sql: shortestIdleTimeLocked correct min comparison
Shinnosuke Sawada [Mon, 17 Aug 2020 11:37:51 +0000 (20:37 +0900)]
database/sql: shortestIdleTimeLocked correct min comparison

When zero or less, maxIdleTime and maxLifetime means unlimited.
Helper function shortestIdleTimeLocked must not return the
minimum of the two until both are verified to be greater
then zero.

Fixes #40841

Change-Id: I1130332baf4ad259cd90c10f4221f5def8510655
Reviewed-on: https://go-review.googlesource.com/c/go/+/248817
Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
4 years agocmd/compile,runtime: skip zero'ing order array for select statements
Cuong Manh Le [Fri, 28 Aug 2020 20:25:21 +0000 (03:25 +0700)]
cmd/compile,runtime: skip zero'ing order array for select statements

The order array was zero initialized by the compiler, but ends up being
overwritten by the runtime anyway.

So let the runtime takes full responsibility for initializing, save us
one instruction per select.

Fixes #40399

Change-Id: Iec1eca27ad7180d4fcb3cc9ef97348206b7fe6b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/251517
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agotest: add test that gccgo failed to compile
Ian Lance Taylor [Thu, 27 Aug 2020 23:31:30 +0000 (16:31 -0700)]
test: add test that gccgo failed to compile

For #38125

Change-Id: Id6ef10d74f0f9dbad2851531e0fe019cd145cf7c
Reviewed-on: https://go-review.googlesource.com/c/go/+/251168
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agotext/template: add CommentNode to template parse tree
Ariel Mashraki [Wed, 22 Apr 2020 19:17:56 +0000 (22:17 +0300)]
text/template: add CommentNode to template parse tree

Fixes #34652

Change-Id: Icf6e3eda593fed826736f34f95a9d66f5450cc98
Reviewed-on: https://go-review.googlesource.com/c/go/+/229398
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agotest: document specifying individual test files as operands
zdjones [Thu, 16 Jul 2020 11:33:25 +0000 (12:33 +0100)]
test: document specifying individual test files as operands

The current command will run this entire set of tests, which takes a
noticeable amount of time. Contributors may wish to run only a subset of
these tests to save time/compute (e.g. when iterating on a CL that
failed tests in that subset). Listing file(s) as operands to the command
will run only those tests.

Change-Id: I1874c43681a594190bc40b61cee0b8d321be73f8
Reviewed-on: https://go-review.googlesource.com/c/go/+/242997
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
4 years agotesting: run a Cleanup registered by a Cleanup
Ian Lance Taylor [Fri, 28 Aug 2020 17:34:16 +0000 (10:34 -0700)]
testing: run a Cleanup registered by a Cleanup

Fixes #41085

Change-Id: Ieafc60cbc8e09f1935d38b1767b084d78dae5cb4
Reviewed-on: https://go-review.googlesource.com/c/go/+/251457
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agogo/types: add tests for conversion of non-constant untyped values
Rob Findley [Fri, 28 Aug 2020 16:10:31 +0000 (12:10 -0400)]
go/types: add tests for conversion of non-constant untyped values

This was fixed by CL 242084. Retroactively add some tests that would
have failed before the fix.

Also, remove some existing duplicate tests.

Change-Id: I95f7a215d4a9651ded6d739f89c574f33f573c60
Reviewed-on: https://go-review.googlesource.com/c/go/+/251397
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agogo/types: factor out usage of implicit type
Rob Findley [Mon, 13 Jul 2020 02:36:34 +0000 (22:36 -0400)]
go/types: factor out usage of implicit type

There was some duplication of logic interpreting the implicit type of
an operand in assignableTo and convertUntyped. Factor out this logic to
a new 'implicitType' function, which returns the implicit type of an
untyped operand when used in a context where a target type is expected.
I believe this resolves some comments about code duplication. There is
other similar code in assignable, assignableTo, and convertUntypes, but
I found it to to be sufficiently semantically distinct to not warrant
factoring out.

Change-Id: I199298a2e58fcf05344318fca0226b460c57867d
Reviewed-on: https://go-review.googlesource.com/c/go/+/242084
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agomath/big: tune addVW/subVW performance on arm64
Xiangdong Ji [Sat, 22 Feb 2020 08:12:54 +0000 (08:12 +0000)]
math/big: tune addVW/subVW performance on arm64

Add an optimization for addVW and subVW over large-sized vectors, it switches
from add/sub with carry to copy the rest of the vector when we are done with
carries. Consistent performance improvement are observed on various arm64
machines.

Add additional tests and benchmarks to increase the test coverage.
TestFunVWExt:
  Testing with various types of input vector, using the result from go-version
  addVW/subVW as golden reference.
BenchmarkAddVWext and BenchmarkSubVWext:
  Benchmarking using input vector having all 1s or all 0s, for evaluating the
  overhead of worst case.

1. Perf. comparison over randomly generated input vectors:

Server 1:
name             old time/op    new time/op    delta
AddVW/1            12.3ns ± 3%    12.0ns ± 0%    -2.60%  (p=0.001 n=10+8)
AddVW/2            12.5ns ± 2%    12.3ns ± 0%    -1.84%  (p=0.001 n=10+8)
AddVW/3            12.6ns ± 2%    12.3ns ± 0%    -1.91%  (p=0.009 n=10+10)
AddVW/4            13.1ns ± 3%    12.7ns ± 0%    -2.98%  (p=0.006 n=10+8)
AddVW/5            14.4ns ± 1%    13.9ns ± 0%    -3.81%  (p=0.000 n=10+10)
AddVW/10           11.7ns ± 0%    11.7ns ± 0%      ~     (all equal)
AddVW/100          47.8ns ± 0%    29.9ns ± 2%   -37.38%  (p=0.000 n=10+9)
AddVW/1000          446ns ± 0%     207ns ± 0%   -53.59%  (p=0.000 n=10+10)
AddVW/10000        4.35µs ± 1%    2.92µs ± 0%   -32.85%  (p=0.000 n=10+10)
AddVW/100000       43.6µs ± 0%    29.7µs ± 0%   -31.92%  (p=0.000 n=8+10)
SubVW/1            12.6ns ± 0%    12.3ns ± 2%    -2.22%  (p=0.000 n=7+10)
SubVW/2            12.7ns ± 0%    12.6ns ± 1%    -0.39%  (p=0.046 n=8+10)
SubVW/3            12.7ns ± 1%    12.6ns ± 1%      ~     (p=0.410 n=10+10)
SubVW/4            13.3ns ± 3%    13.1ns ± 3%      ~     (p=0.072 n=10+10)
SubVW/5            14.2ns ± 0%    14.1ns ± 1%    -0.63%  (p=0.046 n=8+10)
SubVW/10           11.7ns ± 0%    11.7ns ± 0%      ~     (all equal)
SubVW/100          47.8ns ± 0%    33.1ns ±19%   -30.71%  (p=0.000 n=10+10)
SubVW/1000          446ns ± 0%     207ns ± 0%   -53.59%  (p=0.000 n=10+10)
SubVW/10000        4.33µs ± 1%    2.92µs ± 0%   -32.66%  (p=0.000 n=10+6)
SubVW/100000       43.4µs ± 0%    29.6µs ± 0%   -31.90%  (p=0.000 n=10+9)

Server 2:
name             old time/op    new time/op    delta
AddVW/1            5.49ns ± 0%    5.53ns ± 2%     ~     (p=1.000 n=9+10)
AddVW/2            5.96ns ± 2%    5.92ns ± 1%   -0.69%  (p=0.039 n=10+10)
AddVW/3            6.72ns ± 0%    6.73ns ± 0%     ~     (p=0.078 n=10+10)
AddVW/4            7.07ns ± 0%    6.75ns ± 2%   -4.55%  (p=0.000 n=10+10)
AddVW/5            8.14ns ± 0%    8.17ns ± 0%   +0.46%  (p=0.003 n=8+8)
AddVW/10           10.0ns ± 0%    10.1ns ± 1%   +0.70%  (p=0.003 n=10+10)
AddVW/100          43.0ns ± 0%    33.5ns ± 0%  -22.09%  (p=0.000 n=9+9)
AddVW/1000          394ns ± 0%     278ns ± 0%  -29.44%  (p=0.000 n=10+10)
AddVW/10000        4.18µs ± 0%    3.14µs ± 0%  -24.81%  (p=0.000 n=8+8)
AddVW/100000       68.3µs ± 3%    62.1µs ± 5%   -9.13%  (p=0.000 n=10+10)
SubVW/1            5.37ns ± 2%    5.42ns ± 1%     ~     (p=0.990 n=10+10)
SubVW/2            5.89ns ± 0%    5.92ns ± 1%   +0.58%  (p=0.000 n=8+10)
SubVW/3            6.64ns ± 1%    6.82ns ± 3%   +2.63%  (p=0.000 n=9+10)
SubVW/4            7.17ns ± 0%    6.69ns ± 2%   -6.74%  (p=0.000 n=10+9)
SubVW/5            8.22ns ± 0%    8.18ns ± 0%   -0.46%  (p=0.001 n=8+9)
SubVW/10           10.0ns ± 1%    10.1ns ± 1%     ~     (p=0.341 n=10+10)
SubVW/100          43.0ns ± 0%    33.5ns ± 0%  -22.09%  (p=0.000 n=7+10)
SubVW/1000          394ns ± 0%     278ns ± 0%  -29.44%  (p=0.000 n=10+10)
SubVW/10000        4.18µs ± 0%    3.15µs ± 0%  -24.62%  (p=0.000 n=9+9)
SubVW/100000       67.7µs ± 4%    62.4µs ± 2%   -7.92%  (p=0.000 n=10+10)

2. Perf. comparison over input vectors of all 1s or all 0s

Server 1:
name             old time/op    new time/op    delta
AddVWext/1         12.6ns ± 0%    12.0ns ± 0%    -4.76%  (p=0.000 n=6+10)
AddVWext/2         12.7ns ± 0%    12.4ns ± 1%    -2.52%  (p=0.000 n=10+10)
AddVWext/3         12.7ns ± 0%    12.4ns ± 0%    -2.36%  (p=0.000 n=9+7)
AddVWext/4         13.2ns ± 4%    12.7ns ± 0%    -3.71%  (p=0.001 n=10+9)
AddVWext/5         14.6ns ± 0%    13.9ns ± 0%    -4.79%  (p=0.000 n=10+8)
AddVWext/10        11.7ns ± 0%    11.7ns ± 0%      ~     (all equal)
AddVWext/100       47.8ns ± 0%    47.4ns ± 0%    -0.84%  (p=0.000 n=10+10)
AddVWext/1000       446ns ± 0%     399ns ± 0%   -10.54%  (p=0.000 n=10+10)
AddVWext/10000     4.34µs ± 1%    3.90µs ± 0%   -10.12%  (p=0.000 n=10+10)
AddVWext/100000    43.9µs ± 1%    39.4µs ± 0%   -10.18%  (p=0.000 n=10+10)
SubVWext/1         12.6ns ± 0%    12.3ns ± 2%    -2.70%  (p=0.000 n=7+10)
SubVWext/2         12.6ns ± 1%    12.6ns ± 2%      ~     (p=0.234 n=10+10)
SubVWext/3         12.7ns ± 0%    12.6ns ± 2%    -0.71%  (p=0.033 n=10+10)
SubVWext/4         13.4ns ± 0%    13.1ns ± 3%    -2.01%  (p=0.006 n=8+10)
SubVWext/5         14.2ns ± 0%    14.1ns ± 1%    -0.85%  (p=0.003 n=10+10)
SubVWext/10        11.7ns ± 0%    11.7ns ± 0%      ~     (all equal)
SubVWext/100       47.8ns ± 0%    47.4ns ± 0%    -0.84%  (p=0.000 n=10+10)
SubVWext/1000       446ns ± 0%     399ns ± 0%   -10.54%  (p=0.000 n=10+10)
SubVWext/10000     4.33µs ± 1%    3.90µs ± 0%   -10.02%  (p=0.000 n=10+10)
SubVWext/100000    43.5µs ± 0%    39.5µs ± 1%    -9.16%  (p=0.000 n=7+10)

Server 2:
name             old time/op    new time/op    delta
AddVWext/1         5.48ns ± 0%    5.43ns ± 1%   -0.97%  (p=0.000 n=9+9)
AddVWext/2         5.99ns ± 2%    5.93ns ± 1%     ~     (p=0.054 n=10+10)
AddVWext/3         6.74ns ± 0%    6.79ns ± 1%   +0.80%  (p=0.000 n=9+10)
AddVWext/4         7.18ns ± 0%    7.21ns ± 1%   +0.36%  (p=0.034 n=9+10)
AddVWext/5         7.93ns ± 3%    8.18ns ± 0%   +3.18%  (p=0.000 n=10+8)
AddVWext/10        10.0ns ± 0%    10.1ns ± 1%   +0.60%  (p=0.011 n=10+10)
AddVWext/100       43.0ns ± 0%    47.7ns ± 0%  +10.93%  (p=0.000 n=9+10)
AddVWext/1000       394ns ± 0%     399ns ± 0%   +1.27%  (p=0.000 n=10+10)
AddVWext/10000     4.18µs ± 0%    4.50µs ± 0%   +7.73%  (p=0.000 n=9+10)
AddVWext/100000    67.6µs ± 2%    68.4µs ± 3%     ~     (p=0.139 n=9+8)
SubVWext/1         5.46ns ± 1%    5.43ns ± 0%   -0.55%  (p=0.002 n=9+9)
SubVWext/2         5.89ns ± 0%    5.93ns ± 1%   +0.68%  (p=0.000 n=8+10)
SubVWext/3         6.72ns ± 1%    6.79ns ± 1%   +1.07%  (p=0.000 n=10+10)
SubVWext/4         6.98ns ± 1%    7.21ns ± 0%   +3.25%  (p=0.000 n=10+10)
SubVWext/5         8.22ns ± 0%    7.99ns ± 3%   -2.83%  (p=0.000 n=8+10)
SubVWext/10        10.0ns ± 1%    10.1ns ± 1%     ~     (p=0.239 n=10+10)
SubVWext/100       43.0ns ± 0%    47.7ns ± 0%  +10.93%  (p=0.000 n=8+10)
SubVWext/1000       394ns ± 0%     399ns ± 0%   +1.27%  (p=0.000 n=10+10)
SubVWext/10000     4.18µs ± 0%    4.51µs ± 0%   +7.86%  (p=0.000 n=8+8)
SubVWext/100000    68.3µs ± 2%    68.0µs ± 3%     ~     (p=0.515 n=10+8)

Change-Id: I134a5194b8a2deaaebbaa2b771baf72846971d58
Reviewed-on: https://go-review.googlesource.com/c/go/+/229739
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocmd/compile: Install testcases for flag constant Ops
Xiangdong Ji [Thu, 18 Jun 2020 03:36:12 +0000 (03:36 +0000)]
cmd/compile: Install testcases for flag constant Ops

Flag constant Ops on arm and arm64 are under refactoring, this change adds
a couple of testcases that verify the behavior of 'noov' branches.

Updates #39505
Updates #38740
Updates #39303
Change-Id: I493344b52276900cd296c32da494d72932dfc9be
Reviewed-on: https://go-review.googlesource.com/c/go/+/238677
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocmd/asm: Always use go-style arrangement specifiers on ARM64
Xiangdong Ji [Fri, 12 Jun 2020 16:27:08 +0000 (16:27 +0000)]
cmd/asm: Always use go-style arrangement specifiers on ARM64

Fixing several error message and comment texts of the ARM64 assembler
to use arrangement specifiers of Go's assembly style.

Change-Id: Icdbb14fba7aaede40d57d0d754795b050366a1ab
Reviewed-on: https://go-review.googlesource.com/c/go/+/237859
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agomath/big: add a comment for SetMantExp
surechen [Wed, 13 May 2020 09:03:36 +0000 (17:03 +0800)]
math/big: add a comment for SetMantExp

Change-Id: I9ff5d1767cf70648c2251268e5e815944a7cb371
Reviewed-on: https://go-review.googlesource.com/c/go/+/233737
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agodoc: add linux/riscv64 valid combination
Santiago De la Cruz [Thu, 27 Aug 2020 04:00:59 +0000 (04:00 +0000)]
doc: add linux/riscv64 valid combination

Mention valid combination GOOS=linux and GOARCH=riscv64
in the "Installing Go from source" document.

Updates #27532

Change-Id: I8c1406087132f5c82a2eee5dbcda95d53c64d263
GitHub-Last-Rev: ee74ff9517bd428b75400956d5b7f774ff800ae5
GitHub-Pull-Request: golang/go#41063
Reviewed-on: https://go-review.googlesource.com/c/go/+/250997
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
4 years agotesting: restore os.Exit(0) after every call to (*M).Run
Bryan C. Mills [Fri, 28 Aug 2020 01:30:24 +0000 (21:30 -0400)]
testing: restore os.Exit(0) after every call to (*M).Run

cmd/go.TestScript/test_main_twice demonstrates a program that invokes
(*M).Run twice in a row. If we only restore os.Exit(0) in m.afterOnce,
we will fail to restore it after the second run and fail the test
process despite both runs passing.

Updates #29062
Updates #23129

Change-Id: Id22ec68f1708e4583c8dda14a8ba0efae7178b85
Reviewed-on: https://go-review.googlesource.com/c/go/+/251262
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agoRevert "cmd/compile,cmd/asm: simplify recording of branch targets"
Keith Randall [Thu, 27 Aug 2020 23:34:32 +0000 (23:34 +0000)]
Revert "cmd/compile,cmd/asm: simplify recording of branch targets"

This reverts CL 243318.

Reason for revert: Seems to be crashing some builders.

Change-Id: I2ffc59bc5535be60b884b281c8d0eff4647dc756
Reviewed-on: https://go-review.googlesource.com/c/go/+/251169
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile: clean up equality generation
Keith Randall [Mon, 15 Jun 2020 18:08:36 +0000 (11:08 -0700)]
cmd/compile: clean up equality generation

We're using sort.SliceStable, so no need to keep track of indexes as well.

Use a more robust test for whether a node is a call.

Add a test that we're actually reordering comparisons. This test fails
without the alg.go changes in this CL because eqstring uses OCALLFUNC
instead of OCALL for its data comparisons.

Update #8606

Change-Id: Ieeec33434c72e3aa328deb11cc415cfda05632e2
Reviewed-on: https://go-review.googlesource.com/c/go/+/237921
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/go, testing, os: fail test that calls os.Exit(0)
Ian Lance Taylor [Thu, 27 Aug 2020 00:26:05 +0000 (17:26 -0700)]
cmd/go, testing, os: fail test that calls os.Exit(0)

This catches cases where a test calls code that calls os.Exit(0),
thereby skipping all subsequent tests.

Fixes #29062

Change-Id: If9478972f40189e27623557e7141469ca4234d89
Reviewed-on: https://go-review.googlesource.com/c/go/+/250977
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile: invalidate zero-use values during rewrite
Keith Randall [Tue, 30 Jun 2020 22:59:40 +0000 (15:59 -0700)]
cmd/compile: invalidate zero-use values during rewrite

This helps remove uses that aren't needed any more.
That in turn helps other rules with Uses==1 conditions fire.

Update #39918

Change-Id: I68635b675472f1d59e59604e4d34b949a0016533
Reviewed-on: https://go-review.googlesource.com/c/go/+/249463
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
4 years agocmd/compile,cmd/asm: simplify recording of branch targets
Keith Randall [Fri, 17 Jul 2020 19:12:59 +0000 (12:12 -0700)]
cmd/compile,cmd/asm: simplify recording of branch targets

We currently use two fields to store the targets of branches.
Some phases use p.To.Val, some use p.Pcond. Rewrite so that
every branch instruction uses p.To.Val.
p.From.Val is also used in rare instances.
Introduce a Pool link for use by arm/arm64, instead of
repurposing Pcond.

This is a cleanup CL in preparation for some stack frame CLs.

Change-Id: I9055bf0a1d986aff421e47951a1dedc301c846f8
Reviewed-on: https://go-review.googlesource.com/c/go/+/243318
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: framepointers are no longer an experiment - hard code them
Keith Randall [Fri, 21 Aug 2020 18:09:45 +0000 (11:09 -0700)]
runtime: framepointers are no longer an experiment - hard code them

I think they are no longer experimental status. Might as well promote
them to permanent.

Change-Id: Id1259601b3dd2061dd60df86ee48080bfb575d2f
Reviewed-on: https://go-review.googlesource.com/c/go/+/249857
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
4 years agocmd/cgo: ensure GCC does not use ANSI escape sequences in errors
Benjamin Barenblat [Thu, 27 Aug 2020 20:12:18 +0000 (16:12 -0400)]
cmd/cgo: ensure GCC does not use ANSI escape sequences in errors

cgo parses GCC’s error messages to classify C identifiers referenced
from Go programs (are they integer constants? type names?). If GCC tries
to colorize its errors, cgo can’t figure out what GCC is saying. GCC
avoids escape sequences in this scenario by default, but the default
behavior can be overridden in at least two places:

  - The user can set `CGO_COPTS=-fdiagnostics-color`.

  - Whoever compiled GCC can configure GCC itself to always colorize
    output.

The most reliable way to ensure that GCC doesn’t colorize output is to
append `-fdiagnostics-color=never` to the GCC command line; do so.

Fixes #40415

Change-Id: Id4bdf8d92fac8b038340b4264f726e8fe38875b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/248398
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/compile: generate subfic on ppc64
Paul E. Murphy [Thu, 20 Aug 2020 20:06:23 +0000 (15:06 -0500)]
cmd/compile: generate subfic on ppc64

This merges an lis + subf into subfic, and for 32b constants
lwa + subf into oris + ori + subf.

The carry bit is no longer used in code generation, therefore
I think we can clobber it as needed.  Note, lowered borrow/carry
arithmetic is self-contained and thus is not affected.

A few extra rules are added to ensure early transformations to
SUBFCconst don't trip up earlier rules, fold constant operations,
or otherwise simplify lowering.  Likewise, tests are added to
ensure all rules are hit.  Generic constant folding catches
trivial cases, however some lowering rules insert arithmetic
which can introduce new opportunities (e.g BitLen or Slicemask).

I couldn't find a specific benchmark to demonstrate noteworthy
improvements, but this is generating subfic in many of the default
bent test binaries, so we are at least saving a little code space.

Change-Id: Iad7c6e5767eaa9dc24dc1c989bd1c8cfe1982012
Reviewed-on: https://go-review.googlesource.com/c/go/+/249461
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com>
4 years agoruntime: add lock partial order edge (fin -> wbufSpans)
Michael Pratt [Tue, 25 Aug 2020 14:39:02 +0000 (10:39 -0400)]
runtime: add lock partial order edge (fin -> wbufSpans)

runfinq may have write barriers, thus it may need to take wbufSpans on
any write.

Fixes #41021

Change-Id: Ib69e20994b5d7d1526ad53d6ddb5e2e83bf2ed00
Reviewed-on: https://go-review.googlesource.com/c/go/+/250464
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
4 years agocmd/go/internal/modload: address issues missed in CL 244774
Bryan C. Mills [Tue, 25 Aug 2020 01:34:33 +0000 (21:34 -0400)]
cmd/go/internal/modload: address issues missed in CL 244774

For #36460

Change-Id: I5e9a42d64e36679b881e96145833e66cf77b98a5
Reviewed-on: https://go-review.googlesource.com/c/go/+/250338
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agogo/types: remove need to enumerate fixedbugs test files
Rob Findley [Tue, 11 Aug 2020 17:21:14 +0000 (13:21 -0400)]
go/types: remove need to enumerate fixedbugs test files

This is a port of CL 244628, updated to move some existing regression
tests into the fixedbugs directory, and to use subtests. Also,
'TestFixed' is renamed to 'TestFixedBugs'.

Change-Id: I43aac3f75f2bd850567d08e8b008d91aeb717064
Reviewed-on: https://go-review.googlesource.com/c/go/+/247904
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agogo/types: clean up test support code and remove global variables
Rob Findley [Tue, 11 Aug 2020 16:49:40 +0000 (12:49 -0400)]
go/types: clean up test support code and remove global variables

This is a straightforward port of CL 244627.

Change-Id: Ide980957430b35e22a6e22818b0ce9de410988af
Reviewed-on: https://go-review.googlesource.com/c/go/+/247902
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/go/internal/modfetch: do not use mangled version strings to construct module...
Bryan C. Mills [Thu, 27 Aug 2020 03:17:53 +0000 (23:17 -0400)]
cmd/go/internal/modfetch: do not use mangled version strings to construct module.VersionErrors

Better still would be to avoid passing around module.Version instances
with invalid Version strings in the first place, so that any time we
see a module.Version we know that it is actually a version of a module
(and not a structurally-similar datum with something else tacked on to
one of the fields). But that's a bigger cleanup for which I don't
currently have enough bandwidth.

Fixes #41060

Change-Id: I32fba5619105cbf67dd03691064c82b8ebb3ce18
Reviewed-on: https://go-review.googlesource.com/c/go/+/250951
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agogo/types: fix missing Importer for TestBenchmark
Rob Findley [Wed, 19 Aug 2020 02:29:22 +0000 (22:29 -0400)]
go/types: fix missing Importer for TestBenchmark

TestBenchmark is broken due to lack of a Config.Importer, but
unfortunately fails silently due to an unchecked error.

Fix the importer and check the error. Also improve the output to include
allocation stats.

Finally, don't run TestBenchmark on go/types by default. If the
benchmark is being used during a refactoring of go/types itself, results
for go/types will not be comparable.

Change-Id: Ib6bdb6807403b3ec99762f535e2496c94bd9b6e0
Reviewed-on: https://go-review.googlesource.com/c/go/+/249517
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
4 years agocmd/compile: use addressing modes pass on s390x
Michael Munday [Thu, 14 May 2020 14:58:27 +0000 (07:58 -0700)]
cmd/compile: use addressing modes pass on s390x

Add s390x support to the addressing modes pass. This significantly
reduces the number of rules we need to have to handle indexed
addressing modes on s390x.

There are some changes introduced with the new approach. Notably
pointer calculations of the form '(ADD x (ADDconst y [c]))' won't
get fully merged into address fields right now, the constant offset
will remain separate. That is a relatively minor issue though.

file      before    after     Δ       %
addr2line 4120904   4120960   +56     +0.001%
api       4944005   4948765   +4760   +0.096%
asm       4977431   4984335   +6904   +0.139%
buildid   2683760   2683504   -256    -0.010%
cgo       4557976   4558408   +432    +0.009%
compile   19103577  18916634  -186943 -0.979%
cover     4883694   4885054   +1360   +0.028%
dist      3545177   3553689   +8512   +0.240%
doc       3921766   3921518   -248    -0.006%
fix       3295254   3302182   +6928   +0.210%
link      6539222   6540286   +1064   +0.016%
nm        4105085   4107757   +2672   +0.065%
objdump   4546015   4545439   -576    -0.013%
pack      2416661   2415485   -1176   -0.049%
pprof     13267433  13265489  -1944   -0.015%
test2json 2762180   2761996   -184    -0.007%
trace     10145090  10135626  -9464   -0.093%
vet       6772946   6771738   -1208   -0.018%
total     106588176 106418865 -169311 -0.159%

Fixes #37891.

Change-Id: If60d51f31eb2806b011432a6519951b8668cb42f
Reviewed-on: https://go-review.googlesource.com/c/go/+/250958
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agogo/types: add untyped test cases for AssignableTo API
Rob Findley [Mon, 24 Aug 2020 14:43:43 +0000 (10:43 -0400)]
go/types: add untyped test cases for AssignableTo API

The AssignableTo API is specifically for non-constant values, but is
currently called by gopls for constant completions. Add a test to ensure
that we handle this edge case correctly.

Change-Id: I83115cbca2443a783df1c3090b5741260dffb78e
Reviewed-on: https://go-review.googlesource.com/c/go/+/250258
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agogo/types: better error when converting untyped values in assignments
Rob Findley [Sun, 12 Jul 2020 20:57:01 +0000 (16:57 -0400)]
go/types: better error when converting untyped values in assignments

The error returned by convertUntyped is 'cannot convert _ to _', which
can be misleading in contexts where an explicit conversion would be
allowed.

Arguably the error message from convertUntyped should just be 'cannot
use _ as _', as 'convert' has an explicit meaning within the spec.
Making that change caused a large number of test failures, so for now we
just fix this for assignments by interpreting the error.

For #22070

Change-Id: I4eed6f39d1a991e8df7e035ec301d28a05150eb5
Reviewed-on: https://go-review.googlesource.com/c/go/+/242083
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agonet/mail: return error on empty address list
Ian Lance Taylor [Fri, 14 Aug 2020 20:29:07 +0000 (13:29 -0700)]
net/mail: return error on empty address list

This restores the handling accidentally changed in CL 217377.

Fixes #40803
For #36959

Change-Id: If77fbc0c2a1dde4799f760affdfb8dde9bcaf458
Reviewed-on: https://go-review.googlesource.com/c/go/+/248598
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Jeremy Fishman <jfishman@cloudflare.com>
4 years agocmd/compile: remove unused carry related ssa ops in ppc64
Paul E. Murphy [Thu, 20 Aug 2020 20:06:06 +0000 (15:06 -0500)]
cmd/compile: remove unused carry related ssa ops in ppc64

The intermediate SSA opcodes* are no longer generated during the
lowering pass.  The shifting rules have been improved using ISEL.
Therefore, we can remove them and the rules which expand them.

* The removed opcodes are:

  LoweredAdd64Carry
  ADDconstForCarry
  MaskIfNotCarry
  FlagCarryClear
  FlagCarrySet

Change-Id: I1ebe2726ed988f29ed4800c8f57b428f7a214cd0
Reviewed-on: https://go-review.googlesource.com/c/go/+/249462
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
4 years agodoc/go1.14: document json.Umarshal map key support of TextUnmarshaler
ShihCheng Tu [Sun, 14 Jun 2020 16:07:10 +0000 (00:07 +0800)]
doc/go1.14: document json.Umarshal map key support of TextUnmarshaler

Document that json.Unmarshal supports map keys whose underlying
types implement encoding.TextUnmarshaler.

Fixes #38801

Change-Id: Icb9414e9067517531ba0da910bd4a2bb3daace65
Reviewed-on: https://go-review.googlesource.com/c/go/+/237857
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agonet: export ErrClosed
Ian Lance Taylor [Tue, 25 Aug 2020 04:42:20 +0000 (21:42 -0700)]
net: export ErrClosed

This permits programs to reliably detect whether they are using a
closed network connection.

Fixes #4373

Change-Id: Ib4ce8cc82bbb134c4689f0ebc8b9b11bb8b32a22
Reviewed-on: https://go-review.googlesource.com/c/go/+/250357
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Russ Cox <rsc@golang.org>
4 years agocmd/go/internal/modload: reject some bad module paths
Michael Matloob [Mon, 16 Dec 2019 22:18:06 +0000 (17:18 -0500)]
cmd/go/internal/modload: reject some bad module paths

This change rejects module paths that don't conform to
the new checkModulePathLax function, when loading a go.mod
file. The change uses the checkModulePathLax function instead of
CheckPath because there are still many users who are using
unpublished modules with unpublishable paths, and we don't
want to break them all.

Next, before this change, when go mod init is run in GOPATH,
it would try to use the location of the directory within GOPATH
to infer the module path. After this change, it will only use
that inferred module path if it conforms to module.CheckPath.

Change-Id: Idb36d1655cc76aae82671e87ba634609503ad1a2
Reviewed-on: https://go-review.googlesource.com/c/go/+/211597
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agodoc: add module retraction to release notes
Jay Conrod [Wed, 15 Apr 2020 19:11:51 +0000 (15:11 -0400)]
doc: add module retraction to release notes

For #24031

Change-Id: I9bd0905e9aacee4bec3463b7d91f6f0929744752
Reviewed-on: https://go-review.googlesource.com/c/go/+/228384
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go: improve 'go get' handling of retracted versions
Jay Conrod [Wed, 15 Apr 2020 18:52:38 +0000 (14:52 -0400)]
cmd/go: improve 'go get' handling of retracted versions

'go get' will now warn about retracted versions in the build list,
after updating go.mod. The warning instructs users to run
'go get module@latest' to upgrade or downgrade away from the retracted
version.

'go get' now allows users to explicitly request a specific retracted
version.

For #24031

Change-Id: I15fda918dc84258fb35b615dcd33b0f499481bd7
Reviewed-on: https://go-review.googlesource.com/c/go/+/228383
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go: add -retracted flag to 'go list'
Jay Conrod [Wed, 15 Apr 2020 18:42:15 +0000 (14:42 -0400)]
cmd/go: add -retracted flag to 'go list'

The -retracted flag causes 'go list' to load information about
retracted module module versions.

When -retracted is used with -f or -json, the Retracted field is set
to a string containing the reason for the retraction on retracted
module versions. The string is based on comments on the retract
directive. This field is also populated when the -u flag is used.

When -retracted is used with -versions, retracted versions are shown.
Normally, they are omitted.

For #24031

Change-Id: Ic13d516eddffb1b8404e21034f78cecc9896d1b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/228382
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go: add -retract and -dropretract flags to 'go mod edit'
Jay Conrod [Wed, 15 Apr 2020 18:17:08 +0000 (14:17 -0400)]
cmd/go: add -retract and -dropretract flags to 'go mod edit'

'go mod edit' can now add and remove 'retract' directives from go.mod
files.

Also, retractions are now included in the 'go mod edit -json' output.

For #24031

Change-Id: Ife7915e259fa508626d6ec5f786b5c860b489599
Reviewed-on: https://go-review.googlesource.com/c/go/+/228381
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go/internal/modload: support go.mod retract directive
Jay Conrod [Wed, 15 Apr 2020 17:56:09 +0000 (13:56 -0400)]
cmd/go/internal/modload: support go.mod retract directive

The go command now recognizes 'retract' directives in go.mod. A
retract directive may be used by a module author to indicate a
version should not be used. The go command will not automatically
upgrade to a retracted version. Retracted versions will not be
considered when resolving version queries like "latest" that don't
refer to a specific version.

Internally, when the go command resolves a version query, it will find
the highest release version (or pre-release if no release is
available), then it will load retractions from the go.mod file for
that version. Comments on retractions are treated as a rationale and
may appear in error messages. Retractions are only loaded when a query
is resolved, so this should have no impact on performance for most
builds, except when go.mod is incomplete.

For #24031

Change-Id: I17d643b9e03a3445676dbf1a5a351090c6ff6914
Reviewed-on: https://go-review.googlesource.com/c/go/+/228380
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go/internal/modload: refactor version filtering for exclude
Jay Conrod [Wed, 15 Apr 2020 16:08:24 +0000 (12:08 -0400)]
cmd/go/internal/modload: refactor version filtering for exclude

Query and other functions now accept an "allowed" function that
returns an error (previously, the function returned a bool). If the
error is equivalent to ErrDisallowed, it indicates the version is
excluded (or, in a future CL, retracted). This provides predicates a
chance to explain why a version is not allowed.

When a query refers to a specific revision (by version, branch, tag,
or commit name), most callers will not use the Allowed predicate. This
allows commands like 'go list -m' and 'go mod download' to handle
disallowed versions when explicitly requested. 'go get' will reject
excluded versions though.

When a query does not refer to a specific revision (for example,
"latest"), disallowed versions will not be considered.

When an "allowed" predicate returns an error not equivalent to
ErrDisallowed, it may be ignored or returned, depending on the
case. This never happens for excluded versions, but it may happen for
retractions (in a future CL). This indicates a list of retractions
could not be loaded. This frequently happens when offline, and it
shouldn't cause a fatal or warning in most cases.

For #24031

Change-Id: I4df6fb6bd60e3e0259e5b3b4bf71a307b4b32298
Reviewed-on: https://go-review.googlesource.com/c/go/+/228379
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/go: update vendored golang.org/x/mod
Jay Conrod [Mon, 13 Apr 2020 20:33:28 +0000 (16:33 -0400)]
cmd/go: update vendored golang.org/x/mod

This CL vendors go.mod parser changes for the retract directive.

For #24031

Change-Id: Ief19b0eca4c7956eceadc893bb209da7e9ecf22c
Reviewed-on: https://go-review.googlesource.com/c/go/+/228377
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agocmd/go/internal/modload: reject empty go.mod files
Michael Matloob [Mon, 16 Dec 2019 22:18:06 +0000 (17:18 -0500)]
cmd/go/internal/modload: reject empty go.mod files

Don't add a module declaration to a go.mod file when
loading a module. Require a user to call go mod init or to
add the module declaration themselves.

Fixes #35070

Change-Id: If5543580d3c1cfee1fc528eb853b872c4173ca82
Reviewed-on: https://go-review.googlesource.com/c/go/+/234107
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/fix: remove un-used code
Cuong Manh Le [Wed, 26 Aug 2020 16:18:41 +0000 (23:18 +0700)]
cmd/fix: remove un-used code

Change-Id: Ibbd9da6fd1f3219c9c6103d8f858347ab3a21210
Reviewed-on: https://go-review.googlesource.com/c/go/+/250857
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agomath/big: add function example
zhouzhongyuan [Fri, 8 May 2020 10:13:58 +0000 (18:13 +0800)]
math/big: add function example

While reading the source code of the math/big package, I found the SetString function example of float type missing.

Change-Id: Id8c16a58e2e24f9463e8ff38adbc98f8c418ab26
Reviewed-on: https://go-review.googlesource.com/c/go/+/232804
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/internal/objabi: delete doc.go
Cherry Zhang [Sat, 22 Aug 2020 18:31:49 +0000 (14:31 -0400)]
cmd/internal/objabi: delete doc.go

cmd/internal/objabi/doc.go has comments decribing the (old)
object file format. But cmd/internal/objabi has nothing to do
with object files, and never did. Delete.

Move some comment to cmd/internal/goobj, where the (new) object
file format is actually defined, and update to reflect the
current status.

Change-Id: Ied96089df4be35e5d259a572ed60ee00f2cd0d1d
Reviewed-on: https://go-review.googlesource.com/c/go/+/249958
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/compile: apply strong typing to all remaining s390x rewrite rules
Michael Munday [Tue, 25 Aug 2020 13:43:27 +0000 (06:43 -0700)]
cmd/compile: apply strong typing to all remaining s390x rewrite rules

This CL applies strong aux typing to the remaining s390x rewrite
rules in preparation for strong aux typing becoming the default.

Passes toolstash-check on s390x.

Change-Id: Id585b0db492780737818024e1b22b4837435b525
Reviewed-on: https://go-review.googlesource.com/c/go/+/250558
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agomisc/wasm: make wasm_exec more robust against uncommon environments
Richard Musiol [Sat, 15 Aug 2020 19:15:35 +0000 (21:15 +0200)]
misc/wasm: make wasm_exec more robust against uncommon environments

JavaScript environments are quite unpredictable because bundlers add
mocks for compatibility and libraries can polute the global namespace.
Detect more of such situations:

- Add check that require("fs") returns an object.
- Fix check that require("fs") returns an non-empty object.
- Add check that "module" is defined.

Fixes #40730

Change-Id: I2ce65fc7db64bbbb0b60eec79a4cfe5c3fec99c0
Reviewed-on: https://go-review.googlesource.com/c/go/+/248758
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agonet/http: fix detection of Roundtrippers that always error
Michael Fraenkel [Fri, 17 Jul 2020 03:30:12 +0000 (21:30 -0600)]
net/http: fix detection of Roundtrippers that always error

CL 220905 added code to identify alternate transports that always error
by using http2erringRoundTripper. This does not work when the transport
is from another package, e.g., http2.erringRoundTripper.
Expose a new method that allow detection of such a RoundTripper.
Switch to an interface that is both a RoundTripper and can return the
underlying error.

Fixes #40213

Change-Id: I170739857ab9e99dffb5fa55c99b24b23c2f9c54
Reviewed-on: https://go-review.googlesource.com/c/go/+/243258
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agosrc/go.mod, net/http: update bundled and latest golang.org/x/net
Emmanuel T Odeke [Sat, 22 Aug 2020 12:47:06 +0000 (05:47 -0700)]
src/go.mod, net/http: update bundled and latest golang.org/x/net

Updates x/net/http2 to git rev c89045814202410a2d67ec20ecf177ec77ceae7f

    http2: perform connection health check
    https://golang.org/cl/198040 (fixes #31643)

    http2: use ASCII space trimming for parsing Trailer header
    https://golang.org/cl/231437

    all: update golang.org/x/crypto to v0.0.0-20200622213623-75b288015ac9
    https://golang.org/cl/239700 (updates #30965)

    net/http2: fix erringRoundTripper
    https://golang.org/cl/243257 (updates #40213)

also updates the vendored version of golang.org/x/net as per

$ go get golang.org/x/net@c890458142
$ go mod tidy
$ go mod vendor
$ go generate -run bundle std

Change-Id: Iea2473ef086df760144d9656f03a0218eb9da91f
Reviewed-on: https://go-review.googlesource.com/c/go/+/249842
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agogo/types: factor out some methods that compute a single error
Rob Findley [Sun, 12 Jul 2020 20:20:02 +0000 (16:20 -0400)]
go/types: factor out some methods that compute a single error

In order to generate more accurate or informative error messages from
the type checker, it can be helpful to interpret error messages in
context. This is currently achieved in a number of ways:

 + Return a boolean value, and then reverse-engineer the error at the
   callsite (as in representable->representableConst).
 + Return a value causing the error (as in Checker.missingMethod), and
   add the error at the callsite.
 + Pass a "reason" string pointer to capture the error (as in
   Checker.assignableTo), and add the error at the callsite.
 + Pass a "context" string pointer, and use this when writing errors in
   the delegated method.

In all cases, it is the responsibility of whatever code calls
Checker.error* to set the operand mode to invalid.

These methods are used as appropriate, depending on whether multiple
errors are generated, whether additional context is needed, and whether
the mere presence of an error needs to be interpreted at the callsite.
However, this practice has some downsides: the plurality of error
handling techniques can be a barrier to readability and composability.

In this CL, we introduce Yet Another Pattern, with the hope that it can
replace some or all of the existing techniques: factor out side-effect
free functions that evaluate a single error, and add helpers for
recording this error in the Checker.

As a proof of concept this is done for Checker.representable and
Checker.convertUntyped. If the general pattern does not seem appropriate
for replacing some or all of the error-handling techniques listed above,
we should revert to an established technique.

Some internal error APIs are refactored to operate on an error, rather
than a types.Error, with internal error metadata extracted using
errors.As. This seemed to have negligible impact on performance, but we
should be careful about actually wrapping errors: I expect that many
users will expect err to be a types.Error.

Change-Id: Ic5c6edcdc02768cd84e04638fad648934bcf3c17
Reviewed-on: https://go-review.googlesource.com/c/go/+/242082
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agotesting: fix Cleanup race with Logf and Errorf
Michał Łowicki [Sun, 23 Aug 2020 22:53:04 +0000 (23:53 +0100)]
testing: fix Cleanup race with Logf and Errorf

Fixes #40908

Change-Id: I25561a3f18e730a50e6fbf85aa7bd85bf1b73b6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/250078
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agoRevert "net/http: fix data race due to writeLoop goroutine left running"
Bryan C. Mills [Tue, 25 Aug 2020 18:01:27 +0000 (18:01 +0000)]
Revert "net/http: fix data race due to writeLoop goroutine left running"

This reverts CL 232799.

Reason for revert: net/http test is failing on all longtest builders.

Change-Id: I4694e34f35419bab2d0b45fa6d8c3ac2aa1f51a0
Reviewed-on: https://go-review.googlesource.com/c/go/+/250597
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Katie Hockman <katie@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agonet/http: fix data race due to writeLoop goroutine left running
Steven Hartland [Thu, 7 May 2020 21:12:21 +0000 (21:12 +0000)]
net/http: fix data race due to writeLoop goroutine left running

Fix a data race for clients that mutate requests after receiving a
response error which is caused by the writeLoop goroutine left
running, this can be seen on canceled requests.

Fixes #37669

Change-Id: I0e0e4fd63266326b32587d8596456760bf848b13
Reviewed-on: https://go-review.googlesource.com/c/go/+/232799
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

4 years agocrypto/x509: fix duplicate import
Katie Hockman [Tue, 25 Aug 2020 15:30:32 +0000 (11:30 -0400)]
crypto/x509: fix duplicate import

Updates dave/dst#45.

Change-Id: I165e6b3d002407a33908bf90a66ad01f8003b260
Reviewed-on: https://go-review.googlesource.com/c/go/+/250497
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agomath/big: fix TestShiftOverlap for test -count arguments > 1
SparrowLii [Mon, 24 Aug 2020 06:43:32 +0000 (14:43 +0800)]
math/big: fix TestShiftOverlap for test -count arguments > 1

Don't overwrite incoming test data.

The change uses copy instead of assigning statement to avoid this.

Change-Id: Ib907101822d811de5c45145cb9d7961907e212c3
Reviewed-on: https://go-review.googlesource.com/c/go/+/250137
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/compile: report error for unexported name only once
Cuong Manh Le [Mon, 24 Aug 2020 08:23:27 +0000 (15:23 +0700)]
cmd/compile: report error for unexported name only once

Fixes #22921

Change-Id: If29bd962335ac7676ea4f379727db3d55ae1bf8e
Reviewed-on: https://go-review.googlesource.com/c/go/+/250177
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agoencoding/binary: replace constant literals with named constant (cleanup)
Robert Griesemer [Fri, 21 Aug 2020 17:02:44 +0000 (10:02 -0700)]
encoding/binary: replace constant literals with named constant (cleanup)

Follow-up on https://golang.org/cl/247120.

Brought to my attention by Luke McCoy.

Change-Id: I4530c96fb164d23b0ce5311f2cecb1964f2dea74
Reviewed-on: https://go-review.googlesource.com/c/go/+/249837
Reviewed-by: Katie Hockman <katie@golang.org>
4 years agocmd/compile: also check package.function for GOSSAFUNC match
David Chase [Tue, 25 Aug 2020 02:57:01 +0000 (22:57 -0400)]
cmd/compile: also check package.function for GOSSAFUNC match

Old behavior is still enabled because it doesn't hurt to leave
it in and existing users of this feature (there are dozens of
us!) will not be surprised.  Adding this finer control allows
users to avoid writing ssa.html where they can't, shouldn't, or
just don't want to.

Example, both ways:

$ GOSSAFUNC="(*Reader).Reset" go test -c -o ./a compress/gzip
dumped SSA to bytes/ssa.html
dumped SSA to strings/ssa.html
dumped SSA to bufio/ssa.html
dumped SSA to compress/gzip/ssa.html

$ GOSSAFUNC="compress/gzip.(*Reader).Reset" go test -c -o ./a compress/gzip
dumped SSA to compress/gzip/ssa.html

Updates #40919.

Change-Id: I06b77c3c1d326372a32651570b5dd6e56dfb1d7f
Reviewed-on: https://go-review.googlesource.com/c/go/+/250340
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agocmd/link: remove superfluous check in TestIssue34788Android386TLSSequence
Tobias Klauser [Tue, 25 Aug 2020 13:19:50 +0000 (15:19 +0200)]
cmd/link: remove superfluous check in TestIssue34788Android386TLSSequence

err != nil is already checked in the if condition one line above.

Change-Id: If36cdb41016f7be98a65be0a7211d85cd6017f87
Reviewed-on: https://go-review.googlesource.com/c/go/+/250477
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agonet/http: handle Request.URL.RawPath in StripPrefix
Andrew Gerrand [Wed, 13 May 2020 00:39:11 +0000 (10:39 +1000)]
net/http: handle Request.URL.RawPath in StripPrefix

The StripPrefix wrapper strips a prefix string from the request's
URL.Path field, but doesn't touch the RawPath field. This leads to the
confusing situation when StripPrefix handles a request with URL.RawPath
populated (due to some escaped characters in the request path) and the
wrapped request's RawPath contains the prefix but Path does not.

This change modifies StripPrefix to strip the prefix from both Path and
RawPath. If there are escaped characters in the prefix part of the
request URL the stripped handler serves a 404 instead of invoking the
underlying handler with a mismatched Path/RawPath pair.

This is a backward incompatible change for a very small minority of
requests; I would be surprised if anyone is depending on this behavior,
but it is possible. If that's the case, we could make a more
conservative change where the RawPath is trimmed if possible, but when
the prefix contains escaped characters then we don't 404 but rather send
through the invalid Path/RawPath pair as before.

Fixes #24366

Change-Id: I7030b8c183a3dfce307bc0272bba9a18df4cfe08
Reviewed-on: https://go-review.googlesource.com/c/go/+/233637
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agoos: return a *PathError from Readdirnames and Readdir on POSIX platforms
Bryan C. Mills [Wed, 13 May 2020 18:59:29 +0000 (14:59 -0400)]
os: return a *PathError from Readdirnames and Readdir on POSIX platforms

Previously, Readdirnames returned a *PathError on Windows and Plan 9,
but a *SyscallError on POSIX systems.

In contrast, similar methods (such as Stat) return a *PathError on all platforms.

Fixes #38923

Change-Id: I26395905b1e723933f07b792c7aeee7c335949cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/233917
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agogo/build: ignore symlinks to directories when matching source files
Bryan C. Mills [Fri, 26 Jun 2020 15:48:37 +0000 (11:48 -0400)]
go/build: ignore symlinks to directories when matching source files

Fixes #39841

Change-Id: Icbdc37d40e9c10179d6eb704d04482175b139f57
Reviewed-on: https://go-review.googlesource.com/c/go/+/240120
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/compile: don't allow go:notinheap on the heap or stack
Keith Randall [Sat, 22 Aug 2020 03:20:12 +0000 (20:20 -0700)]
cmd/compile: don't allow go:notinheap on the heap or stack

Right now we just prevent such types from being on the heap. This CL
makes it so they cannot appear on the stack either. The distinction
between heap and stack is pretty vague at the language level (e.g. it
is affected by -N), and we don't need the flexibility anyway.

Once go:notinheap types cannot be in either place, we don't need to
consider pointers to such types to be pointers, at least according to
the garbage collector and stack copying. (This is the big win of this
CL, in my opinion.)

The distinction between HasPointers and HasHeapPointer no longer
exists. There is only HasPointers.

This CL is cleanup before possible use of go:notinheap to fix #40954.

Update #13386

Change-Id: Ibd895aadf001c0385078a6d4809c3f374991231a
Reviewed-on: https://go-review.googlesource.com/c/go/+/249917
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
4 years agocmd/go/internal/par: add Queue as a simpler alternative to Work
Bryan C. Mills [Mon, 10 Aug 2020 19:11:07 +0000 (15:11 -0400)]
cmd/go/internal/par: add Queue as a simpler alternative to Work

par.Work performs two different tasks: deduplicating work (a task
which overlaps with par.Cache), and executing limited active work in
parallel. It also requires the caller to re-invoke Do whenever the
workqueue transititions from empty to non-empty.

The new par.Queue only performs the second of those two tasks, and
presents a simpler API: it starts and stops its own goroutines as
needed (indicating its idle state via a channel), rather than
expecting the caller to drive the transitions explicitly.

For #36460

Change-Id: I5c38657dda63ab55718497467d05d41744ff59f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/247766
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go: add baseline test cases for non-lazy module loading
Bryan C. Mills [Fri, 21 Feb 2020 17:23:47 +0000 (12:23 -0500)]
cmd/go: add baseline test cases for non-lazy module loading

For #36460
For #40799

Change-Id: Id55934cc4d66743a4087b4c2644b6c3b95e7d2ce
Reviewed-on: https://go-review.googlesource.com/c/go/+/222341
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agocmd/go/internal/modload: cache parsed go.mod files globally
Bryan C. Mills [Thu, 23 Jul 2020 04:45:27 +0000 (00:45 -0400)]
cmd/go/internal/modload: cache parsed go.mod files globally

Previously they were cached per mvsReqs instance. However, the
contents of the go.mod file of a given dependency version can only
vary if the 'replace' directives that apply to that version have
changed, and the only time we change 'replace' directives is in 'go
mod edit' (which does not care about the build list or MVS).

This not only simplifies the mvsReqs implementation, but also makes
more of the underlying logic independent of mvsReqs.

For #36460

Change-Id: Ieac20c2fcd56f64d847ac8a1b40f9361ece78663
Reviewed-on: https://go-review.googlesource.com/c/go/+/244774
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/modload: cache the Go language version for each module globally
Bryan C. Mills [Tue, 21 Jul 2020 20:50:59 +0000 (16:50 -0400)]
cmd/go/internal/modload: cache the Go language version for each module globally

Previously, this cache was a member of the (ephemeral) modload.loader
struct. However, the Go language version for a given module version
does not vary based on the build list, the set of loaded packages, the
build tags in use, the meaning of the "all" pattern, or anything else
that can be configured for an instance of the package loader. The map
containing that information is therefore not appropriate as a field of
the (configurable, package-list-dependent) loader struct.

The Go language version mapping could, in theory, be read from the
go.mod file in the module cache (or replacement directory) every time
it is needed: this map is just a cache, and as such it belongs
alongside the other caches and indexes in the modload package, which
are currently found in modfile.go.

We may want to do the same sort of global caching for the mapping from
each module.Version to its list of direct requirements (which are
similarly idempotent), but for now that is left for a future change.

For #36460
For #36876

Change-Id: I90ac176ffea97f30c47d6540c3dfb874dc9cfa4f
Reviewed-on: https://go-review.googlesource.com/c/go/+/244078
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agocmd/go/internal/mvs: indicate the actual version when printing a mismatched ModuleError
Bryan C. Mills [Wed, 5 Aug 2020 03:53:01 +0000 (23:53 -0400)]
cmd/go/internal/mvs: indicate the actual version when printing a mismatched ModuleError

Previously, we suppressed the module version annotation if the last
error in the stack was a *module.ModuleError, regardless of its path.
However, if the error is for a replacement module, that produces a
confusing error message: the error is attributed to the last module in
the error path, but actually originates in the replacement (which is
not otherwise indicated).

Now, we print both the original and the replacement modules when they
differ, which may add some unfortunate redundancy in the output but at
least doesn't drop the very relevant information about replacements.

Fixes #35039

Change-Id: I631a7398033602b1bd5656150a4fad4945a87ade
Reviewed-on: https://go-review.googlesource.com/c/go/+/247765
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/mvs: export a NewBuildListError function
Bryan C. Mills [Wed, 5 Aug 2020 03:27:18 +0000 (23:27 -0400)]
cmd/go/internal/mvs: export a NewBuildListError function

Also factor out BuildListError to a separate file.

For #36460

Change-Id: Ibd1143893b09a2bbef659bea1e8c5dd35184a7ef
Reviewed-on: https://go-review.googlesource.com/c/go/+/247764
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/mvs: reverse the order of BuildListError.stack
Bryan C. Mills [Fri, 24 Jul 2020 02:24:04 +0000 (22:24 -0400)]
cmd/go/internal/mvs: reverse the order of BuildListError.stack

When we print the stack from a BuildListError, we print the main
module first and the error last. That was the opposite of the order in
which in was stored in memory, leading to (arguably) more complex code
and (definitely) my own inability to reason about the contents of the
slice.

For now, it's still more convenient to construct the stack reversed,
so we do that and then reverse it before packing it into the error.

For #36460

Change-Id: I6312fb67b2ad9bf9b64071fe829854833208bad7
Reviewed-on: https://go-review.googlesource.com/c/go/+/244759
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/modload: drop requirements on excluded versions
Bryan C. Mills [Fri, 24 Jul 2020 21:43:55 +0000 (17:43 -0400)]
cmd/go/internal/modload: drop requirements on excluded versions

Previously, when we encountered an excluded version in any module's
requirements, we would resolve it to the next higher version.
Unfortunately, the meaning of “the next higher version” can change
over time.

Moreover, users who use 'exclude' directives normally either already
require some higher version (using the 'exclude' directive to prune
out invalid requirements from some intermediate version), or already
require some lower version (using the 'exclude' directive to prevent
'go get -u' from upgrading to a known-bad version). In both of these
cases, resolving an upgrade for the excluded version is needless work
even in the best case: it adds work for the 'go' command when there is
already a perfectly usable selected version of the module in the
requirement graph.

Instead, we now interpret the 'exclude' directive as dropping all
references to the excluded version.

This implements the approach described in
https://golang.org/issue/36465#issuecomment-572694990.

Fixes #36465
Updates #36460

Change-Id: Ibf0187daced417b4cc23b97125826778658e4b0f
Reviewed-on: https://go-review.googlesource.com/c/go/+/244773
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/modload: add a "v" prefix to the indexed go version
Bryan C. Mills [Fri, 13 Mar 2020 20:46:51 +0000 (16:46 -0400)]
cmd/go/internal/modload: add a "v" prefix to the indexed go version

This allows semver-based comparisons of the version without additional allocations.

Also comment on the reason for the loops that iterate over modFile instead.

(I was reading the vendor code in order to add the lazy-loading version check,
and this section was a bit unclear to me.)

For #36460

Change-Id: I11559d81ffb4eba0e4e10e6fa3c01990b11f9180
Reviewed-on: https://go-review.googlesource.com/c/go/+/240622
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agosync: delete dirty keys inside Map.LoadAndDelete
Changkun Ou [Mon, 24 Aug 2020 11:45:27 +0000 (13:45 +0200)]
sync: delete dirty keys inside Map.LoadAndDelete

Fixes #40999

Change-Id: Ie32427e5cb5ed512b976b554850f50be156ce9f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/250197
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>