Mikio Hara [Tue, 25 Feb 2014 14:02:19 +0000 (23:02 +0900)]
syscall: add mmap test
This CL adds a test that calls Mmap and Munmap through Syscall9
as the canary that detects assembly fragment breakage. For now
there is no package test that uses Syscall9 in the standard
library across all Unix-like systems.
Note that the package runtime owns its assembly fragments, so
this canary never works for runtime breakage.
Adam Langley [Mon, 24 Feb 2014 22:57:51 +0000 (17:57 -0500)]
crypto/tls: pick ECDHE curves based on server preference.
Currently an ECDHE handshake uses the client's curve preference. This
generally means that we use P-521. However, P-521's strength is
mismatched with the rest of the cipher suite in most cases and we have
a fast, constant-time implementation of P-256.
With this change, Go servers will use P-256 where the client supports
it although that can be overridden in the Config.
Rémy Oudompheng [Mon, 24 Feb 2014 18:51:59 +0000 (19:51 +0100)]
cmd/gc: fix walkcompare bugs.
Revision c0e0467635ec (cmd/gc: return canonical Node* from temp)
exposed original nodes of temporaries, allowing callers to mutate
their types.
In walkcompare a temporary could be typed as ideal because of
this. Additionnally, assignment of a comparison result to
a custom boolean type was broken.
Lucio De Re [Mon, 24 Feb 2014 18:48:06 +0000 (19:48 +0100)]
cmd/go: Plan 9 compatible "env" output
Fixes the output of go env so that variables can be set
more accurately when using Plan 9's rc shell. Specifically,
GOPATH may have multiple components and the current
representation is plain wrong. In practice, we probably
ought to change os. Getenv to produce the right result, but
that requires considerably more thought.
Dmitriy Vyukov [Mon, 24 Feb 2014 16:53:50 +0000 (20:53 +0400)]
runtime: fix heap memory corruption
With concurrent sweeping finc if modified by runfinq and queuefinalizer concurrently.
Fixes crashes like this one:
http://build.golang.org/log/6ad7b59ef2e93e3c9347eabfb4c4bd66df58fd5a
Fixes #7324.
Update #7396
Dmitriy Vyukov [Mon, 24 Feb 2014 16:53:20 +0000 (20:53 +0400)]
runtime: fix potential memory corruption
Reinforce the guarantee that MSpan_EnsureSwept actually ensures that the span is swept.
I have not observed crashes related to this, but I do not see why it can't crash as well.
Dmitriy Vyukov [Mon, 24 Feb 2014 16:23:03 +0000 (20:23 +0400)]
runtime: fix double symbol definition
runfinqv is already defined the same way on line 271.
There may also be something to fix in compiler/linker wrt diagnostics.
Fixes #7375.
Jay Weisskopf [Mon, 24 Feb 2014 15:57:46 +0000 (10:57 -0500)]
runtime: use monotonic clock for timers (linux/386, linux/amd64)
This lays the groundwork for making Go robust when the system's
calendar time jumps around. All input values to the runtimeTimer
struct now use the runtime clock as a common reference point.
This affects net.Conn.Set[Read|Write]Deadline(), time.Sleep(),
time.Timer, etc. Under normal conditions, behavior is unchanged.
Each platform and architecture's implementation of runtime·nanotime()
should be modified to use a monotonic system clock when possible.
Platforms/architectures modified and tested with monotonic clock:
linux/x86 - clock_gettime(CLOCK_MONOTONIC)
Shane Hansen [Mon, 24 Feb 2014 15:13:27 +0000 (10:13 -0500)]
gdb: Add partial python3 + go1.2 support to runtime-gdb.py
Update #6963 Fixes pretty printing maps and updates
functions for interacting with $len(). goroutine $n bt
remains not working. Tested on gdb using python 2 and 3.
Fixes #7052
Update #6963
Fixes #6698
Dmitriy Vyukov [Mon, 24 Feb 2014 14:12:46 +0000 (18:12 +0400)]
runtime/race: fix finalizer tests
After "runtime: combine small NoScan allocations" finalizers
for small objects run more non deterministically.
TestRaceFin episodically fails on my darwin/amd64.
LGTM=khr
R=golang-codereviews, khr, dave
CC=golang-codereviews
https://golang.org/cl/56970043
Mikio Hara [Mon, 24 Feb 2014 05:41:10 +0000 (14:41 +0900)]
syscall: consolidate test cases for Unix-like systems
As per request from minux in CL 61520049, this CL consolidates
existing test cases for Unix-like systems into one file except
Linux-specific credential test.
Adam Langley [Fri, 21 Feb 2014 20:56:41 +0000 (15:56 -0500)]
crypto/tls: enforce that either ServerName or InsecureSkipVerify be given.
crypto/tls has two functions for creating a client connection: Dial,
which most users are expected to use, and Client, which is the
lower-level API.
Dial does what you expect: it gives you a secure connection to the host
that you specify and the majority of users of crypto/tls appear to work
fine with it.
Client gives more control but needs more care. Specifically, if it
wasn't given a server name in the tls.Config then it didn't check that
the server's certificates match any hostname - because it doesn't have
one to check against. It was assumed that users of the low-level API
call VerifyHostname on the certificate themselves if they didn't supply
a hostname.
A review of the uses of Client both within Google and in a couple of
external libraries has shown that nearly all of them got this wrong.
Thus, this change enforces that either a ServerName or
InsecureSkipVerify is given. This does not affect tls.Dial.
See discussion at https://groups.google.com/d/msg/golang-nuts/4vnt7NdLvVU/b1SJ4u0ikb0J.
Russ Cox [Thu, 20 Feb 2014 21:18:05 +0000 (16:18 -0500)]
runtime/debug: add SetPanicOnFault
SetPanicOnFault allows recovery from unexpected memory faults.
This can be useful if you are using a memory-mapped file
or probing the address space of the current program.
Russ Cox [Thu, 20 Feb 2014 20:58:47 +0000 (15:58 -0500)]
runtime: use goc2c as much as possible
Package runtime's C functions written to be called from Go
started out written in C using carefully constructed argument
lists and the FLUSH macro to write a result back to memory.
For some functions, the appropriate parameter list ended up
being architecture-dependent due to differences in alignment,
so we added 'goc2c', which takes a .goc file containing Go func
declarations but C bodies, rewrites the Go func declaration to
equivalent C declarations for the target architecture, adds the
needed FLUSH statements, and writes out an equivalent C file.
That C file is compiled as part of package runtime.
Native Client's x86-64 support introduces the most complex
alignment rules yet, breaking many functions that could until
now be portably written in C. Using goc2c for those avoids the
breakage.
Separately, Keith's work on emitting stack information from
the C compiler would require the hand-written functions
to add #pragmas specifying how many arguments are result
parameters. Using goc2c for those avoids maintaining #pragmas.
For both reasons, use goc2c for as many Go-called C functions
as possible.
This CL is a replay of the bulk of CL 15400047 and CL 15790043,
both of which were reviewed as part of the NaCl port and are
checked in to the NaCl branch. This CL is part of bringing the
NaCl code into the main tree.
No new code here, just reformatting and occasional movement
into .h files.
LGTM=r
R=dave, alex.brainman, r
CC=golang-codereviews
https://golang.org/cl/65220044
Russ Cox [Thu, 20 Feb 2014 20:50:30 +0000 (15:50 -0500)]
cmd/pack: fix match
Match used len(ar.files) == 0 to mean "match everything"
but it also deleted matched things from the list, so once you
had matched everything you asked for, match returned true
for whatever was left in the archive too.
Concretely, if you have an archive containing f1, f2, then
pack t foo.a f1
would match f1 and then, because len(ar.files) == 0 after
deleting f1 from the match list, also match f2.
Avoid the problem by recording explicitly whether match
matches everything.
cmd/ld: fix off-by-one error in DWARF .debug_line transcription
The liblink refactor changed the DWARF .debug_line flow control. The mapping was off by one pcline entry. The fix here preserves pc until it can be compared to pcline.pc.
Sample dwarfdump .debug_line output for main.main from the program in issue 7351, before liblink (correct):
Rob Pike [Thu, 20 Feb 2014 01:17:36 +0000 (17:17 -0800)]
syscall: terminate error string in exec package on Plan 9
Try to prevent messages like this:
'./pack' file does not exist����������������������������������������������������������������������������������������������������
TBR=adonovan
Rob Pike [Wed, 19 Feb 2014 23:01:50 +0000 (15:01 -0800)]
cmd/pack: dump output of command of "go env" command in test
Get more information to help understand build failure on Plan 9.
Also Windows.
(TestHello is failing because GOCHAR does not appear in output.
What does?)
Precisestack makes stack collection completely precise,
in the sense that there are no "used and not set" errors
in the collection of stack frames, no times where the collector
reads a pointer from a stack word that has not actually been
initialized with a pointer (possibly a nil pointer) in that function.
The most important part is interfaces: precisestack means
that if reading an interface value, the interface value is guaranteed
to be initialized, meaning that the type word can be relied
upon to be either nil or a valid interface type word describing
the data word.
This requires additional zeroing of certain values on the stack
on entry, which right now costs about 5% overall execution
time in all.bash. That cost will come down before Go 1.3
(issue 7345).
There are at least two known garbage collector bugs right now,
issues 7343 and 7344. The first happens even without precisestack.
The second I have only seen with precisestack, but that does not
mean that precisestack is what causes it. In fact it is very difficult
to explain by what precisestack does directly. Precisestack may
be exacerbating an existing problem. Both of those issues are
marked for Go 1.3 as well.
The reasons for enabling precisestack now are to give it more
time to soak and because the copying stack work depends on it.
Russ Cox [Wed, 19 Feb 2014 22:08:44 +0000 (17:08 -0500)]
cmd/pack: add 'c' command to create archive
When Go 1.3 is released, this will keep existing
Go 1.2 build scripts that use 'go tool pack grc' working.
For efficiency, such scripts should be changed to
use 6g -pack instead, but keeping the old behavior
available enables a more graceful transition.
Rob Pike [Wed, 19 Feb 2014 20:50:50 +0000 (15:50 -0500)]
cmd/gc: fix printf format in typecheck.c
There are probably more of these, but bound and len are 64 bits so use %lld
in message about array index out of bounds.
Fixes the 386 build.
Adam Langley [Wed, 19 Feb 2014 16:17:09 +0000 (11:17 -0500)]
crypto/tls: improve documentation for ServerName.
Users of the low-level, Client function are frequenctly missing the
fact that, unless they pass a ServerName to the TLS connection then it
cannot verify the certificates against any name.
This change makes it clear that at least one of InsecureSkipVerify and
ServerName should always be set.
Russ Cox [Wed, 19 Feb 2014 15:01:15 +0000 (10:01 -0500)]
cmd/go: skip writing dwarf debug info for ephemeral binaries
Update #6853
For an ephemeral binary - one created, run, and then deleted -
there is no need to write dwarf debug information, since the
binary will not be used with gdb. In this case, instruct the linker
not to spend time and disk space generating the debug information
by passing the -w flag to the linker.
Omitting dwarf information reduces the size of most binaries by 25%.
We may be more aggressive about this in the future.
LGTM=bradfitz, r
R=r, bradfitz
CC=golang-codereviews
https://golang.org/cl/65890043
Russ Cox [Wed, 19 Feb 2014 15:00:44 +0000 (10:00 -0500)]
cmd/ld: drop gcargs, gclocals symbols from symbol table
Update #6853
Every function now has a gcargs and gclocals symbol
holding associated garbage collection information.
Put them all in the same meta-symbol as the go.func data
and then drop individual entries from symbol table.
Removing gcargs and gclocals reduces the size of a
typical binary by 10%.
Russ Cox [Wed, 19 Feb 2014 04:41:15 +0000 (23:41 -0500)]
cmd/ld: remove Plan 9 symbol table
Update #6853
Nothing reads the Plan 9 symbol table anymore.
The last holdout was 'go tool nm', but since being rewritten in Go
it uses the standard symbol table for the binary format
(ELF, Mach-O, PE) instead.
Removing the Plan 9 symbol table saves ~15% disk space
on most binaries.
Two supporting changes included in this CL:
debug/gosym: use Go 1.2 pclntab to synthesize func-only
symbol table when there is no Plan 9 symbol table
debug/elf, debug/macho, debug/pe: ignore final EOF from ReadAt
Dominik Honnef [Wed, 19 Feb 2014 03:23:55 +0000 (22:23 -0500)]
misc/emacs: add support for ff-find-other-file
c-mode classically uses ff-find-other-file to toggle between headers
and implementation. For Go it seemingly makes sense to jump between
implementation and test.
While there's no enforced mapping of file names for tests, the mapping
in this CL seems to be very common at least throughout the standard
library, and ff-find-other-file fails gracefully when the mapping
doesn't apply.
Marcel van Lohuizen [Tue, 18 Feb 2014 19:12:59 +0000 (20:12 +0100)]
unicode: upgrade to Unicode 6.3.0
This is a relatively minor change.
This does not result in changes to go.text/unicode/norm. The go.text
packages will therefore be relatively unaffected. It does make the
way for an upgrade to CLDR 24, though.
The tests of all.bash pass, as well as the tests in go.text after
this update.
««« original CL description
cmd/gc, runtime: enable precisestack by default
Precisestack makes stack collection completely precise,
in the sense that there are no "used and not set" errors
in the collection of stack frames, no times where the collector
reads a pointer from a stack word that has not actually been
initialized with a pointer (possibly a nil pointer) in that function.
The most important part is interfaces: precisestack means
that if reading an interface value, the interface value is guaranteed
to be initialized, meaning that the type word can be relied
upon to be either nil or a valid interface type word describing
the data word.
This requires additional zeroing of certain values on the stack
on entry, which right now costs about 5% overall execution
time in all.bash. That cost will come down before Go 1.3
(issue 7345).
There are at least two known garbage collector bugs right now,
issues 7343 and 7344. The first happens even without precisestack.
The second I have only seen with precisestack, but that does not
mean that precisestack is what causes it. In fact it is very difficult
to explain by what precisestack does directly. Precisestack may
be exacerbating an existing problem. Both of those issues are
marked for Go 1.3 as well.
The reasons for enabling precisestack now are to give it more
time to soak and because the copying stack work depends on it.
Russ Cox [Tue, 18 Feb 2014 01:12:40 +0000 (20:12 -0500)]
cmd/gc, runtime: enable precisestack by default
Precisestack makes stack collection completely precise,
in the sense that there are no "used and not set" errors
in the collection of stack frames, no times where the collector
reads a pointer from a stack word that has not actually been
initialized with a pointer (possibly a nil pointer) in that function.
The most important part is interfaces: precisestack means
that if reading an interface value, the interface value is guaranteed
to be initialized, meaning that the type word can be relied
upon to be either nil or a valid interface type word describing
the data word.
This requires additional zeroing of certain values on the stack
on entry, which right now costs about 5% overall execution
time in all.bash. That cost will come down before Go 1.3
(issue 7345).
There are at least two known garbage collector bugs right now,
issues 7343 and 7344. The first happens even without precisestack.
The second I have only seen with precisestack, but that does not
mean that precisestack is what causes it. In fact it is very difficult
to explain by what precisestack does directly. Precisestack may
be exacerbating an existing problem. Both of those issues are
marked for Go 1.3 as well.
The reasons for enabling precisestack now are to give it more
time to soak and because the copying stack work depends on it.
Russ Cox [Tue, 18 Feb 2014 01:11:53 +0000 (20:11 -0500)]
runtime: clear f, arg to avoid leak in timerproc
I have seen this cause leaks where not all objects in a sync.Pool
would be reclaimed during the sync package tests.
I found it while debugging the '0 of 100 finalized' failure we are
seeing on arm, but it seems not to be the root cause for that one.
Dmitriy Vyukov [Mon, 17 Feb 2014 02:29:56 +0000 (06:29 +0400)]
testing: ease writing parallel benchmarks
Add b.RunParallel function that captures parallel benchmark boilerplate:
creates worker goroutines, joins worker goroutines, distributes work
among them in an efficient way, auto-tunes grain size.
Fixes #7090.
Russ Cox [Sun, 16 Feb 2014 01:01:15 +0000 (20:01 -0500)]
cmd/gc: record &x[0] as taking address of x, if x is an array
Not recording the address being taken was causing
the liveness analysis not to preserve x in the absence
of direct references to x, which in turn was making the
net test fail with GOGC=0.
In addition to the test, this fixes a bug wherein
GOGC=0 go test -short net
crashed if liveness analysis was in use (like at tip, not like Go 1.2).