Cherry Zhang [Mon, 8 Jun 2020 22:38:59 +0000 (18:38 -0400)]
[dev.link] cmd/compile, cmd/link: remove dead methods if type is not used in interface
Currently, a method of a reachable type is live if it matches a
method of a reachable interface. In fact, we only need to retain
the method if the type is actually converted to an interface. If
the type is never converted to an interface, there is no way to
call the method through an interface method call (but the type
descriptor could still be used, e.g. in calling
runtime.newobject).
A type can be used in an interface in two ways:
- directly converted to interface. (Any interface counts, as it
is possible to convert one interface to another.)
- obtained by reflection from a related type (e.g. obtaining an
interface of T from []T).
For the former, we let the compiler emit a marker on the type
descriptor symbol when it is converted to an interface. In the
linker, we only need to check methods of marked types.
For the latter, when the linker visits a marked type, it needs to
visit all its "child" types as marked (i.e. potentially could be
converted to interface).
Matthew Dempsky [Wed, 10 Jun 2020 19:47:23 +0000 (12:47 -0700)]
go/types: rename UsesCgo to go115UsesCgo
This API and functionality was added late in the Go 1.15 release
cycle, and use within gopls has revealed some shortcomings. It's
possible (but not decided) that we'll want a different API long-term,
so for now this CL renames UsesCgo to a non-exported name to avoid
long-term commitment under the Go 1 compat guarantee.
Michael Munday [Tue, 9 Jun 2020 10:17:17 +0000 (03:17 -0700)]
cmd/compile: always tighten and de-duplicate tuple selectors
The scheduler assumes two special invariants that apply to tuple
selectors (Select0 and Select1 ops):
1. There is only one tuple selector of each type per generator.
2. Tuple selectors and generators reside in the same block.
Prior to this CL the assumption was that these invariants would
only be broken by the CSE pass. The CSE pass therefore contained
code to move and de-duplicate selectors to fix these invariants.
However it is also possible to write relatively basic optimization
rules that cause these invariants to be broken. For example:
(A (Select0 (B))) -> (Select1 (B))
This rule could result in the newly added selector (Select1) being
in a different block to the tuple generator (see issue #38356). It
could also result in duplicate selectors if this rule matches
multiple times for the same tuple generator (see issue #39472).
The CSE pass will 'fix' these invariants. However it will only do
so when optimizations are enabled (since disabling optimizations
disables the CSE pass).
This CL moves the CSE tuple selector fixup code into its own pass
and makes it mandatory even when optimizations are disabled. This
allows tuple selectors to be treated like normal ops for most of
the compilation pipeline until after the new pass has run, at which
point we need to be careful to maintain the invariant again.
Fixes #39472.
Change-Id: Ia3f79e09d9c65ac95f897ce37e967ee1258a080b
Reviewed-on: https://go-review.googlesource.com/c/go/+/237118
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Bryan C. Mills [Wed, 10 Jun 2020 02:00:18 +0000 (22:00 -0400)]
run.bat: do not unset GOROOT_FINAL before running tests
This removes the same logic from run.bat that was removed from
cmd/dist in CL 236819.
The duplicated logic was removed from run.bash and run.rc in CL 6531,
but that part of run.bat was apparently missed (and not noticed
because its effect was redundant).
Also fix a path-separator bug in cmd/addr2line.TestAddr2Line that was
exposed as a result.
Fixes #39478
Updates #39385
Change-Id: I00054966cf92ef92a03681bf23de7f45f46fbb5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/237359
Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Roland Shoemaker [Fri, 15 May 2020 19:49:04 +0000 (12:49 -0700)]
crypto/tls: restore OCSP and SCTs during session resumption
Restore previously sent SCTs and stapled OCSP response during session
resumption for both TLS 1.2 and 1.3. This behavior is somewhat
complicated for TLS 1.2 as SCTs are sent during the server hello,
so they override what is saved in ClientSessionState. It is likely
that if the server is sending a different set of SCTs there is probably
a reason for doing so, such as a log being retired, or SCT validation
requirements changing, so it makes sense to defer to the server in
that case.
Brad Fitzpatrick [Tue, 9 Jun 2020 16:11:11 +0000 (09:11 -0700)]
runtime: fix typo in comment (object -> objects)
Change-Id: I2af1f9dcd1a9609681e58ab07e73e6d7a5f8a12b
Reviewed-on: https://go-review.googlesource.com/c/go/+/237160 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Xiangdong Ji [Mon, 1 Jun 2020 11:01:14 +0000 (11:01 +0000)]
cmd/compile: ARM comparisons with 0 incorrect on overflow
Some ARM rewriting rules convert 'comparing to zero' conditions of if
statements to a simplified version utilizing CMN and CMP instructions to
branch over condition flags, in order to save one Add or Sub caculation.
Such optimizations lead to wrong branching in case an overflow/underflow
occurs when executing CMN or CMP.
Fix the issue by introducing new block opcodes that don't honor the
overflow/underflow flag:
Block-Op Meaning ARM condition codes
1. LTnoov less than MI
2. GEnoov greater than or equal PL
3. LEnoov less than or equal MI || EQ
4. GTnoov greater than NEQ & PL
The patch also adds a few test cases to cover scenarios that are specific
to ARM and fine-tunes the code generation tests for 'x-const'.
For more details please refer to the previous fix on 64-bit ARM:
https://go-review.googlesource.com/c/go/+/233097
Go1 perf, 'old' is the non-optimized version, that is removing all concerned
rewriting rules.
Dmitri Shuralyov [Tue, 9 Jun 2020 14:26:09 +0000 (14:26 +0000)]
doc/go1.15: remove TODO in minor library changes section
The minor changes to the library section has been populated
with TODOs for individual packages using relnote in CL 235757,
and they've been resolved in the following CLs.
We will look things over as part of finishing touches on
the release notes, but this TODO is resolved for beta 1.
For #37419.
Change-Id: I942f81a957fe8df8f630b4406ca29f73602d080a
Reviewed-on: https://go-review.googlesource.com/c/go/+/237157 Reviewed-by: Alexander Rakoczy <alex@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org>
Bryan C. Mills [Fri, 5 Jun 2020 20:08:08 +0000 (16:08 -0400)]
cmd/dist: do not unset GOROOT_FINAL prior to running tests
Also do not unset it by default in the tests for cmd/go.
GOROOT_FINAL affects the GOROOT value embedded in binaries,
such as 'cmd/cgo'. If its value changes and a build command
is performed that depends on one of those binaries, the binary
would be spuriously rebuilt.
Instead, only unset it in the specific tests that make assumptions
about the GOROOT paths embedded in specific compiled binaries.
That may cause those tests to do a little extra rebuilding when
GOROOT_FINAL is set, but that little bit of extra rebuilding
seems preferable to spuriously-stale binaries.
Fixes #39385
Change-Id: I7c87b1519bb5bcff64babf1505fd1033ffa4f4fb
Reviewed-on: https://go-review.googlesource.com/c/go/+/236819
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Russ Cox [Mon, 8 Jun 2020 16:39:56 +0000 (12:39 -0400)]
all: avoid awkward wording from CL 236857
CL 236857 removed all uses of whitelist/blacklist, which is great.
But it substituted awkward phrasing using allowlist/blocklist,
especially as verbs or participles. This CL uses more standard English,
like "allow the function" or "blocked functions" instead of
"allowlist the function" or "blocklisted functions".
Change-Id: I9106a2fdbd62751c4cbda3a77181358a8a6d0f13
Reviewed-on: https://go-review.googlesource.com/c/go/+/236917
Run-TryBot: Russ Cox <rsc@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Bryan C. Mills [Sat, 6 Jun 2020 02:35:18 +0000 (22:35 -0400)]
cmd/go: remove a bogus assertion in mod_convert_dep
The removed line assumed that the script's WORK directory is not a
child of any directory containing version-control metadata.
While that assumption does hold in most cases, it does not hold when,
for example, $TMPDIR is $HOME/tmp and $HOME/.git/config exists.
A similar situation may or may not arise when using
golang.org/x/build/cmd/release. Either way, the assertion is incorrect
and was interfering with local testing for #39385.
Updates #39385
Fixes #39431
Change-Id: I67813d7ce455aa9b56a6eace6eddebf48d0f7fa6
Reviewed-on: https://go-review.googlesource.com/c/go/+/236818 Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Austin Clements [Sat, 6 Jun 2020 02:01:25 +0000 (22:01 -0400)]
runtime: always mark span when marking an object
The page sweeper depends on spans being marked if any object in the
span is marked, but currently only greyobject does this.
gcmarknewobject and wbBufFlush1 also mark objects, but neither set
span marks. As a result, if there are live objects on a span, but
they're all marked via allocation or write barriers, then the span
itself won't be marked and the page reclaimer will free the span,
ultimately leading to memory corruption when the memory for those live
allocations gets reused.
Fix this by making gcmarknewobject and wbBufFlush1 also mark pages.
No test because I have no idea how to reliably (or even unreliably)
trigger this.
Fixes #39432.
Performance is a wash or very slightly worse. I benchmarked the
gcmarknewobject and wbBufFlush1 changes independently and both showed
a slight performance improvement, so I'm going to call this noise.
Dmitri Shuralyov [Thu, 4 Jun 2020 23:20:01 +0000 (19:20 -0400)]
cmd/internal/moddeps: don't skip directories if there are unusual files
Previously, if there was a non-directory file with the name vendor or
testdata in the Go source tree, it was possible for some directories
to be skipped by filepath.Walk performed in findGorootModules.
As unusual and unlikely as such non-directory files are, it's better
to ensure all directories are visited, and all modules in the GOROOT
source tree are found.
This increases confidence that tests relying on findGorootModule
will not have unexpected false negatives.
For #36851.
For #36907.
Change-Id: I468e80d8f57119e2c72d546b3fd1e23c31fd6e6c
Reviewed-on: https://go-review.googlesource.com/c/go/+/236600
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Dmitri Shuralyov [Thu, 4 Jun 2020 04:35:09 +0000 (00:35 -0400)]
all: treat all files as binary, but check in .bat with CRLF
This is a followup to CL 96495.
It should be simpler and more robust to achieve .bat files having
CRLF line endings by treating it as a binary file, like all other
files, and checking it in with the desired CRLF line endings.
A test is used to check the entire Go tree, short of directories
starting with "." and named "testdata", for any .bat files that
have anything other than strict CRLF line endings. This will help
catch any accidental modifications to existing .bat files or check
ins of new .bat files.
Importantly, this is compatible with how Gerrit serves .tar.gz files,
making it so that CRLF line endings are preserved.
The Go project is supported on many different environments, some of
which may have limited git implementations available, or none at all.
Relying on fewer git features and special rules makes it easier to
have confidence in the exact content of all files. Additionally, Go
development started in Subversion, moved to Perforce, then Mercurial,
and now uses Git.¹ Reducing its reliance on git-specific features will
help if there will be another transition in the project's future.
There are only 5 .bat files in the entire Go source tree, so a new one
being added is a rare event, and we prefer to do things in Go instead.
We still have the option of improving the experience for developers by
adding a pre-commit converter for .bat files to the git-codereview tool.
Filippo Valsorda [Sun, 7 Jun 2020 00:59:12 +0000 (20:59 -0400)]
all: replace usages of whitelist/blacklist and master/slave
There's been plenty of discussion on the usage of these terms in tech.
I'm not trying to have yet another debate. It's clear that there are
people who are hurt by them and who are made to feel unwelcome by their
use due not to technical reasons but to their historical and social
context. That's simply enough reason to replace them.
Anyway, allowlist and blocklist are more self-explanatory than whitelist
and blacklist, so this change has negative cost.
Didn't change vendored, bundled, and minified files. Nearly all changes
are tests or comments, with a couple renames in cmd/link and cmd/oldlink
which are extremely safe. This should be fine to land during the freeze
without even asking for an exception.
Austin Clements [Fri, 5 Jun 2020 13:46:21 +0000 (09:46 -0400)]
doc/go1.15: rationalize runtime sections
Use the "Core library -> runtime" section for changes that affect the
runtime package API and use the top-level "Runtime" section for
package-independent behavior changes. Also, move the one change that's
really about os (and net) into the "os" package section and reword it
to be more accurate.
Updates #37419.
Change-Id: I32896b039f29ac67308badd0d0b36e8c6e39f64f
Reviewed-on: https://go-review.googlesource.com/c/go/+/236718 Reviewed-by: Michael Knyszek <mknyszek@google.com>
Carlos Amedee [Fri, 5 Jun 2020 22:12:22 +0000 (18:12 -0400)]
doc/go1.15: remove TODO intended for the Core library section
The TODO was added durring the initial creation of the document.
In the current location, it makes it seem like the tzdata documents
are incomplete when they are complete. It is understood that the
entire Core library section will be a work in progress until the release.
For #37419
Change-Id: Ic857eb0ec2583781c701985ea62e519e9d940090
Reviewed-on: https://go-review.googlesource.com/c/go/+/236760 Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Rebecca Stambler [Fri, 22 May 2020 03:58:39 +0000 (23:58 -0400)]
doc: update contribution guide to make it friendlier for x/ repos
The current contributor documentation is tailored towards contributors
to golang/go, but we have a number of increasingly popular x/ repos.
In this CL, I tried to generalize the language to make it apply to any
repository.
Also, I fixed an old link I noticed in editors.html.
Change-Id: Id9d8e448262ed8c3a67f49be5d554ca29df9d3c1
Reviewed-on: https://go-review.googlesource.com/c/go/+/234899
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Roland Shoemaker [Mon, 11 May 2020 23:21:54 +0000 (16:21 -0700)]
crypto/x509: match cgo and Apple behavior in domain fallback of macOS roots
This change makes the direct call darwin loadSystemRoots implementation
match the existing cgo implementation, which in turn _mostly_ matches
the Apple implementation. The main change here is that when
SecTrustSettingsCopyTrustSettings the error is ignored, and can either
cause a fallback to check admin trust settings, or cause the
certificate to be marked kSecTrustSettingsResultUnspecified.
As well as updating the implementation to match the cgo one, this
change also updates the documentation of how the fallbacks work and
how they match the Apple implementations. References are made to the
Apple source where appropriate. This change does not update the
existing comments in the cgo implementation, since the goal is to
delete that code once the direct call implementation is matured.
Cherry Zhang [Wed, 3 Jun 2020 23:34:29 +0000 (19:34 -0400)]
[dev.link] cmd/link: rewrite heap algorithm
Instead of using container/heap package, implement a simple
specialized heap algorithm for the work queue in the deadcode
pass, to avoid allocations and function pointer calls.
Linking cmd/compile,
name old time/op new time/op delta
Deadcode_GC 59.8ms ± 4% 42.2ms ± 4% -29.45% (p=0.008 n=5+5)
name old alloc/op new alloc/op delta
Deadcode_GC 3.53MB ± 0% 2.10MB ± 0% -40.57% (p=0.008 n=5+5)
name old allocs/op new allocs/op delta
Deadcode_GC 187k ± 0% 8k ± 0% -95.48% (p=0.008 n=5+5)
Change-Id: Ibb21801d5b8e4a7eaf429856702e02720cd1772f
Reviewed-on: https://go-review.googlesource.com/c/go/+/236565
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Cherry Zhang [Wed, 3 Jun 2020 17:28:18 +0000 (13:28 -0400)]
[dev.link] cmd/link: compare type descriptors for interface satisfaction analysis
Currently, in the deadcode pass, when checking whether a defined
method satisfies an interface, it compares the string
representation of the defined method and the interface method.
In fact, it can simply compare the method name and the type
descriptor (as we do in runtime). Make it so.
Change-Id: Ideb2b2410e5eedcd20ac31e3af41f5499fc90225
Reviewed-on: https://go-review.googlesource.com/c/go/+/236564
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Austin Clements [Thu, 4 Jun 2020 15:16:43 +0000 (11:16 -0400)]
syscall: hide internal comment from Syscall documentation
There's a comment on the Syscall function that's supposed to be an
internal implementation note, but since it's not separated from the
function definition, it appears in godoc. Add a blank line to prevent
this.
Cherry Zhang [Mon, 1 Jun 2020 23:36:16 +0000 (19:36 -0400)]
[dev.link] cmd/link: use fingerprint as package hash
Now the compiler-generated fingerprint is a hash of the export
data. We don't need to hash it ourselves in the linker. And the
linker doesn't need to read export data at all.
Fixes #33820.
Change-Id: I54bf3ebfd0f0c72aa43a352d7b2e0575dd62970d
Reviewed-on: https://go-review.googlesource.com/c/go/+/236119
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Cherry Zhang [Mon, 1 Jun 2020 22:53:58 +0000 (18:53 -0400)]
[dev.link] cmd/compile: use hash of export data as fingerprint
Currently, the compiler generates a fingerprint for each package,
which is used by the linker for index consistency check.
When building plugin or shared object, currently the linker also
generates a hash, by hashing the export data. At run time, when
a package is referenced by multiple DSOs, this hash is compared
to ensure consistency.
It would be good if we can unify this two hashes. This way, the
linker doesn't need to read the export data (which is intended
for the compiler only, and is not always available for the
linker). The export data hash is sufficient for both purposes.
It is consistent with the current hash geneated by the linker.
And the export data includes indices for exported symbols, so its
hash can be used to catch index mismatches.
Updates #33820.
Change-Id: I2bc0d74930746f54c683a10dfd695d50ea3f5a38
Reviewed-on: https://go-review.googlesource.com/c/go/+/236118
Run-TryBot: Cherry Zhang <cherryyz@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
joshuabezaleel [Fri, 7 Feb 2020 10:36:26 +0000 (17:36 +0700)]
io/ioutil: update WriteFile to clarify it does not change permissions if the file exists.
The existing documentation of WriteFile does not make it clear for
non-native English speakers that it will not change the permissions if
the file already exists before.
Fixes #35711
Change-Id: If861c3e3700957fc9ac3d5313351c57d399d3f58
Reviewed-on: https://go-review.googlesource.com/c/go/+/218417 Reviewed-by: Rob Pike <r@golang.org>
David Chase [Wed, 3 Jun 2020 18:21:18 +0000 (14:21 -0400)]
runtime: make runtime-gdb.py tolerant of creatively-named gdb versions
"Fedora" and "Red Hat" are not numbers, it turns out.
Don't rely on version numbers, instead use a regexp to
handle variation across the 2 patterns thus far observed
for gdb-generated Go type names.
Cherry Zhang [Tue, 2 Jun 2020 21:45:57 +0000 (17:45 -0400)]
cmd/internal/goobj2: add referenced symbol names to object file
Currently, for symbols defined in other packages and referenced
by index, we don't record its name in the object file, as the
linker doesn't need the name, only the index. As a consequence,
tools like objdump and nm also don't know the referenced symbol
names and cannot dump it properly.
This CL adds referenced symbol names to the object file. So the
object file is self-contained. And tools can retrieve referenced
symbol names properly.
Tools now should work as good for new object files as for old
object files.
Fixes #38875.
Change-Id: I16c685c1fd83273ab1faef474e19acf4af46396f
Reviewed-on: https://go-review.googlesource.com/c/go/+/236168 Reviewed-by: Austin Clements <austin@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
Cherry Zhang [Tue, 2 Jun 2020 18:52:16 +0000 (14:52 -0400)]
Revert "cmd/internal/goobj: add index to symbol name for indexed symbols"
This reverts CL 229246.
For new indexed object files, in CL 229246 we added symbol index
to tools (nm, objdump) output. This affects external tools that
parse those outputs. And the added index doesn't look very nice.
In this release we take it out. For future releases we may
introduce a flag to tools (nm, objdump) and optionally dump the
symbol index.
For refererenced (not defined) indexed symbols, currently the
symbol is still referenced only by index, not by name. The next
CL will make the object file self-contained, so tools can dump
the symbol names properly (as before).
For #38875.
Change-Id: I07375e85a8e826e15c82fa452d11f0eaf8535a00
Reviewed-on: https://go-review.googlesource.com/c/go/+/236167 Reviewed-by: Than McIntosh <thanm@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org>
Than McIntosh [Tue, 2 Jun 2020 21:15:35 +0000 (17:15 -0400)]
doc/go1.15: add release notes for plugin
Add a blurb to the release notes mentioning that the
linker now supports DWARF generation for -buildmode=plugin,
and that plugin builds work now for freebsd/amd64.
David Chase [Tue, 2 Jun 2020 21:53:47 +0000 (17:53 -0400)]
runtime: repair gdb printing fix for 7.12, 8.{1,2,3}.1, 9.2
Hand-verified for listed gdb versions. Gdb (apparently)
changed the way it names certain Go types, and this change
broke the pretty-printer-activating code in runtime-gdb.py
runtime-gdb_test.go now checks channel, map, string, and slice
printing unconditionally (i.e., no opt-out for old versions).
Updates #39368.
Change-Id: I98d72e1291c66bd40d970990e1a377ff2ed0c5d2
Reviewed-on: https://go-review.googlesource.com/c/go/+/236164
Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Austin Clements [Wed, 3 Jun 2020 01:51:39 +0000 (21:51 -0400)]
doc/go1.15: runtime release notes
Change-Id: Ie37e993e840df2c063dee98fa3f6eca8e8713ca3
Reviewed-on: https://go-review.googlesource.com/c/go/+/236177 Reviewed-by: Michael Knyszek <mknyszek@google.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
fanzha02 [Wed, 3 Jun 2020 02:33:03 +0000 (10:33 +0800)]
cmd/internal/obj/arm64: fix typos in document
The current document mismatches Go syntax loads a signed-byte
instruction "MOVB" with GNU syntax loads an 64bit double-word
instruction "ldr". This is just a typo in the document, the
assembler has the correct encoding. This patch fix this error.
Than McIntosh [Fri, 29 May 2020 20:01:08 +0000 (16:01 -0400)]
cmd/{compile,link}: fix problem with DWARF end_sequence ops
During DWARF line table emission in the linker, prior to issuing a
DW_LNE_end_sequence op to mark the end of the line table for a
compilation unit, advance the PC to produce an address beyond the last
text address in the unit (this is required by the DWARF standard).
Because of the way that GDB interprets end-sequence ops, we were
effectively losing the last row in the line table for each unit, which
degraded the debugging experience.
This problem has been around for a while, but has surfaced recently
due to changes in line table generation. Prior to Go 1.14, the DWARF
line table was emitted entirely in the linker, and a single monolithic
line table was created for each Go package (including functions from
assembly). In 1.14 we moved to having the compiler emit line table
fragments for each function, and having the linker stitch together the
fragments. As part of this change we moved to a model in which each
"go tool compile/asm" output has its own DWARF line table instance,
meaning that there are many more "end sequence" ops, which made the
problem more visible.
Fixes #38192.
Change-Id: Ic29e2f6e0ac952360c81fcba5268ad70b2b44184
Reviewed-on: https://go-review.googlesource.com/c/go/+/235739
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com> Reviewed-by: Jeremy Faller <jeremy@golang.org>
Ian Lance Taylor [Mon, 4 May 2020 23:19:16 +0000 (16:19 -0700)]
runtime: steal timers from running P's
Previously we did not steal timers from running P's, because that P
should be responsible for running its own timers. However, if the P
is running a CPU-bound G, this can cause measurable delays in running
ready timers. Also, in CL 214185 we avoided taking the timer lock of a P
with no ready timers, which reduces the chances of timer lock contention.
So, if we can't find any ready timers on sleeping P's, try stealing
them from running P's.
Fixes #38860
Change-Id: I0bf1d5dc56258838bdacccbf89493524e23d7fed
Reviewed-on: https://go-review.googlesource.com/c/go/+/232199
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Andrew G. Morgan [Tue, 2 Jun 2020 20:13:54 +0000 (13:13 -0700)]
cmd/go: add -Wl,-wrap,symbol to linker flag whitelist
This is needed for cgo support for libpsx.
Fixes: #39361
Change-Id: I500f5614ea4b82b085322af1f1ffeb1f55270a05
Reviewed-on: https://go-review.googlesource.com/c/go/+/236139
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Jean de Klerk [Mon, 1 Jun 2020 18:19:05 +0000 (12:19 -0600)]
testing: capture testname on --- PASS and --- FAIL lines
This fixes an issue raised at https://github.com/golang/go/issues/38458#issuecomment-635617670
in which --- PASS and --- FAIL lines would not trigger --- CONT lines
of other tests.
David Golden [Thu, 23 Apr 2020 01:16:49 +0000 (21:16 -0400)]
time: note that formats may parse invalid strings
The existing documentation for time format constants doesn't mention
that they may parse technically-invalid strings, such as single-digit
hours when a two-digit hour is required by a specification. This commit
adds a short warning note to that effect.
Fixes #37616
Change-Id: I6e5e12bd42dc368f8ca542b4c0527a2b7d30acaf
Reviewed-on: https://go-review.googlesource.com/c/go/+/229460
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Roland Shoemaker [Wed, 13 May 2020 01:49:34 +0000 (18:49 -0700)]
crypto/x509: document support for leading periods in DNS constraints
This change adds a comment to the Verify documentation that indicates
that you can use URI and email style name constraints with a leading
period for DNS names (and explains what they do). This behavior is
not standards compliant, but matches the community application of
RFC 5280, so it makes sense to document it.
Dan Scales [Tue, 2 Jun 2020 15:50:54 +0000 (08:50 -0700)]
runtime: add three new partial orders for lock ranking
Two are associated with the new sysmon rank: lockRankSysmon -> lockRankFin
(https://build.golang.org/log/07b0b8ee6ec9421d83699a1d850d9938390fb996)
and one I encountered during testing, lockRankSysmon -> lockRankWbufSpans
Richard Miller [Tue, 2 Jun 2020 09:34:09 +0000 (10:34 +0100)]
internal/poll: add mutex to prevent SetDeadline race in Plan 9
There are data races on fd.[rw]aio and fd.[rw]timedout when Read/Write
is called on a polled fd concurrently with SetDeadline (see #38769).
Adding a mutex around accesses to each pair (read and write) prevents
the race, which was causing deadlocks in net/http tests on the builders.
Updates #38769.
Change-Id: I31719b3c9a664e81a775cda583cff31c0da946c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/235820
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David du Colombier <0intro@gmail.com>
Cherry Zhang [Mon, 1 Jun 2020 22:06:45 +0000 (18:06 -0400)]
misc/cgo/testplugin: fix typo in comment
Change-Id: I7d1a5f6936505dff8f765541b5102dcbcd6ae835
Reviewed-on: https://go-review.googlesource.com/c/go/+/235924 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Elias Naur [Mon, 1 Jun 2020 09:47:23 +0000 (11:47 +0200)]
doc: document the new Cgo EGLConfig special case
Change-Id: I7ae5eaa974b85eac421a0b1f79cb734a0fe44e72
Reviewed-on: https://go-review.googlesource.com/c/go/+/235818 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Elias Naur [Mon, 1 Jun 2020 09:58:54 +0000 (11:58 +0200)]
doc: document new Android default linker
Change-Id: I3557f6726afe325db79b2c972d107b3bcc103b8f
Reviewed-on: https://go-review.googlesource.com/c/go/+/235819 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Cherry Zhang [Fri, 29 May 2020 05:59:09 +0000 (01:59 -0400)]
cmd/link: flush file mapping before unmapping
Call FlushViewOfFile before unmapping the output file, for extra
safety. The documentation says the function does not wait for
the data to be written to disk, so it should be cheap.
Fixes #38440.
Change-Id: I05352f15d9305e6e7086a002f61802f74036b710
Reviewed-on: https://go-review.googlesource.com/c/go/+/235639 Reviewed-by: Austin Clements <austin@google.com> Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Jay Conrod [Thu, 14 May 2020 19:03:26 +0000 (15:03 -0400)]
cmd/go/internal/modload: document mvsReqs.Max
The version "" denotes the main module, which has no version. The
mvs.Reqs interface documentation hints this is allowed, but it's not
obvious from the implementation in modload.mvsReqs.Max.
Also, replace a related TODO with a comment in mvs.Downgrade.
Fixes #39042
Change-Id: I11e10908c9b3d8c2283eaa5c04bd8e1b936851fd
Reviewed-on: https://go-review.googlesource.com/c/go/+/234003
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Daniel Nephin [Mon, 1 Jun 2020 16:11:34 +0000 (16:11 +0000)]
cmd/test2json: attribute output to the correct test
When printing regular test output check the indentation of the output, and use
the report stack to find the appropriate test name for that output.
This change includes a whitespace change to some golden test files. The
indentation of tests was changed in CL 113177
from tabs to spaces. The golden files have been updated to match the new
output format. The tabs in the golden files cause problems because the indentation check
looks for 4 spaces.
Richard Musiol [Sat, 25 Apr 2020 16:53:53 +0000 (18:53 +0200)]
runtime: fix race condition between timer and event handler
This change fixes a race condition between beforeIdle waking up the
innermost event handler and a timer causing a different goroutine to
wake up at the exact same moment. This messes up the wasm event handling
and leads to memory corruption. The solution is to make beforeIdle
return the goroutine that must run next and have findrunnable pick
this goroutine without considering timers again.