Russ Cox [Wed, 17 Feb 2016 15:39:13 +0000 (10:39 -0500)]
doc: remove DRAFT tags from go1.6.html
Go 1.6 is soon (but not yet).
Fixes #14301.
Change-Id: I85e329b643adcb5d4fa680c5333fbc1f928d4d9d
Reviewed-on: https://go-review.googlesource.com/19550 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>
Russ Cox [Wed, 17 Feb 2016 15:41:03 +0000 (10:41 -0500)]
net/http: update bundle command
This is the bundle command's new usage and new output header,
after CL 19428.
Actually running this command would work but would bring in
a newer x/net/http2 that we don't want yet.
Change-Id: Ic6082ca00102a2df1f7632eebf9aca41fdcdb444
Reviewed-on: https://go-review.googlesource.com/19551 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>
Russ Cox [Wed, 17 Feb 2016 15:46:30 +0000 (10:46 -0500)]
doc: mention multi-change branches in contribute.html
Fixes #13651.
Change-Id: I1d21b49e2b5bc6c507eb084d6d2553e5a9c607cf
Reviewed-on: https://go-review.googlesource.com/19552 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>
Austin Clements [Tue, 16 Feb 2016 19:42:24 +0000 (14:42 -0500)]
runtime: fix deadlock in TestCrashDumpsAllThreads
TestCrashDumpsAllThreads carefully sets the number of Ps to one
greater than the number of non-preemptible loops it starts so that the
main goroutine can continue to run (necessary because of #10958).
However, if GC starts, it can take over that one spare P and lock up
the system while waiting for the non-preemptible loops, causing the
test to eventually time out. This deadlock is easily reproducible if
you run the runtime test with GOGC=1.
Fix this by forcing GOGC=off when running this test.
Russ Cox [Tue, 16 Feb 2016 17:09:05 +0000 (12:09 -0500)]
cmd/compile: print struct tags in var decl in inlined func body
This bug was introduced in golang.org/cl/18217,
while trying to fix #13777.
Originally I wanted to just disable inlining for the case
being handled incorrectly, but it's fairly difficult to detect
and much easier just to fix. Since the case being handled
incorrectly was inlined correctly in Go 1.5, not inlining it
would also be somewhat of a regression.
So just fix it.
Test case copied from Ian's CL 19520.
The mistake to worry about in this CL would be relaxing
the condition too much (we now print the note more often
than we did yesterday). To confirm that we'd catch this mistake,
I checked that changing (!fmtbody || !t.Funarg) to (true) does
cause fixedbugs/issue13777.go to fail. And putting it back
to what is written in this CL makes that test pass again
as well as the new fixedbugs/issue14331.go.
So I believe that the new condition is correct for both constraints.
Fixes #14331.
Change-Id: I91f75a4d5d07c53af5caea1855c780d9874b8df6
Reviewed-on: https://go-review.googlesource.com/19514
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Alex Brainman [Wed, 10 Feb 2016 23:15:58 +0000 (10:15 +1100)]
net: make TestInterfaceAddrsWithNetsh more robust
TestInterfaceAddrsWithNetsh invokes Windows netsh command passing
it a particular interface name. This approach somehow does not work
on some computers (see issue for details). Change that to call netsh
without specifying any interface name. This provides output for all
interfaces available. So we can achieve same goal parsing this output.
Also makes test faster because we only need to invoke netsh once.
Austin Clements [Fri, 12 Feb 2016 15:33:51 +0000 (10:33 -0500)]
runtime: show panics in traceback
We used to include panic calls in tracebacks; however, when
runtime.panic was renamed to runtime.gopanic in the conversion of the
runtime to Go, we missed the special case in showframe that includes
panic calls even though they're in package runtime.
Fix the function name check in showframe (and, while we're here, fix
the other check for "runtime.panic" in runtime/pprof). Since the
"runtime.gopanic" name doesn't match what users call panic and hence
isn't very user-friendly, make traceback rewrite it to just "panic".
Austin Clements [Sun, 14 Feb 2016 20:09:21 +0000 (15:09 -0500)]
doc: tweak Go 1.6 GC release notes
Go 1.6 significantly improves pause times for large heaps, but it
improves them in many other situations as well, such as when goroutine
churn is high, allocation rate is high, or when there are many
finalizers. Hence, make the statement about pause times a bit more
general.
Austin Clements [Mon, 15 Feb 2016 04:53:19 +0000 (23:53 -0500)]
doc: fix non-possessive "'s"
Currently we use "Section's" as the plural of the debug/elf Section
struct. Change this to "Sections" because it's not possessive and
doesn't seem to fall in to any special cases were the apostrophe is
acceptable.
Keith Randall [Fri, 12 Feb 2016 18:07:36 +0000 (10:07 -0800)]
cmd/compile: add write barrier to type switch
Type switches need write barriers if the written-to
variable is heap allocated.
For the added needwritebarrier call, the right arg doesn't
really matter, I just pass something that will never disqualify
the write barrier. The left arg is the one that matters.
Fixes #14306
Change-Id: Ic2754167cce062064ea2eeac2944ea4f77cc9c3b
Reviewed-on: https://go-review.googlesource.com/19481 Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Fri, 12 Feb 2016 02:09:34 +0000 (18:09 -0800)]
runtime: return errno value from Solaris mmap as expected
The code in mem_bsd.go expects that when mmap fails it will return a
positive errno value. This fixes the Solaris implementation of mmap to
work as expected.
Change-Id: Id1c34a9b916e8dc955ced90ea2f4af8321d92265
Reviewed-on: https://go-review.googlesource.com/19477
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Ian Lance Taylor [Thu, 11 Feb 2016 05:46:51 +0000 (21:46 -0800)]
runtime: fix errno sign for some mmap and mincore cases
The caller of mmap expects it to return a positive errno value, but the
linux-arm64 and nacl-386 system calls returned a negative errno value.
Correct them to negate the errno value.
The caller of mincore expects it to return a negative errno value (yes,
this is inconsistent), but the linux-mips64x and linux-ppc64x system
call returned a positive errno value. Correct them to negate the errno
value.
Add a test that mmap returns errno with the correct sign. Brad added a
test for mincore's errno value in https://golang.org/cl/19457.
Fixes #14297.
Change-Id: I2b93f32e679bd1eae1c9aef9ae7bcf0ba39521b5
Reviewed-on: https://go-review.googlesource.com/19455 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Brad Fitzpatrick [Wed, 10 Feb 2016 19:58:44 +0000 (19:58 +0000)]
net/http/httptest: make Server.CloseClientConnections wait for conns to close
httptest.Server was rewritten during Go 1.6, but
CloseClientConnections was accidentally made async in the rewrite and
not caught due to lack of tests.
Robert Griesemer [Tue, 9 Feb 2016 20:58:35 +0000 (12:58 -0800)]
go/internal/gcimporter: interpret relative srcDir relative to cwd
1) go/types.dir: Correctly return "." if there is no path.
2) go/internal/gcimporter.FindPkg: work-around for build.Import
(build.Import doesn't produce expected result if srcDir is
relative). See also issue 14282.
Volker Dobler [Tue, 9 Feb 2016 13:39:52 +0000 (14:39 +0100)]
net: make getmac based tests on windows more robust
The Windows 7 getmac command may report the physical address of an adapter
as "Disabled" or "N/A". Handle these two cases to make the tests more
robust when building on Windows with manually disabled adapters or turned
off hardware.
Addresses issue #14130.
Change-Id: I0c2f8554b4b6810568e4e60ed53857599401f296
Reviewed-on: https://go-review.googlesource.com/19411 Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Brad Fitzpatrick [Tue, 9 Feb 2016 20:05:10 +0000 (20:05 +0000)]
net/http: be more conservative about enabling http2 on Transports
For now, don't enable http2 when Transport.TLSConfig != nil.
See background in #14275.
Also don't enable http2 when ExpectContinueTimeout is specified for
now, in case somebody depends on that functionality. (It is not yet
implemented in http2, and was only just added to net/http too in Go
1.6, so nobody would be setting it yet).
Russ Cox [Tue, 9 Feb 2016 04:46:27 +0000 (23:46 -0500)]
cmd/go: use GOPATH order for compile -I and link -L options
Given GOPATH=p1:p2 and source code of just the right form,
the go command could previously end up invoking the compiler
with -I p2 -I p1 or the linker with -L p2 -L p1, so that
compiled packages in p2 incorrectly shadowed packages in p1.
If foo were in both p1 and p2 and the compilation of bar
were such that the -I and -L options were inverted in this way,
then
GOPATH=p2 go install foo
GOPATH=p1:p2 go install bar
would get the p2 copy of foo instead of the (expected) p1 copy of foo.
This manifested in real usage in a few different ways, but in all
the root cause was that the -I or -L option sequence did not
match GOPATH.
Make it match GOPATH.
Fixes #14176 (second report).
Fixes #14192.
Related but less common issue #14271 not fixed.
Change-Id: I9c0f69042bb2bf92c9fc370535da2c60a1187d30
Reviewed-on: https://go-review.googlesource.com/19385 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Russ Cox [Tue, 9 Feb 2016 03:50:17 +0000 (22:50 -0500)]
cmd/go: silence standard imports non-standard error for non-existent import target
This error only affects the compilation of the standard library,
but I discovered that if you import "notexist" from the standard
library then you get both an error about notexist not existing
and an error about notexist being a non-standard package
(because the non-existant package is in fact not a standard package).
Silence the second error.
Brad Fitzpatrick [Mon, 8 Feb 2016 23:23:36 +0000 (23:23 +0000)]
net/http: make ListenAndServeTLS treat GetCertificate as a set cert too
ListenAndServeTLS doesn't require cert and key file names if the
server's TLSConfig has a cert configured. This code was never updated
when the GetCertificate hook was added to *tls.Config, however.
Brad Fitzpatrick [Mon, 8 Feb 2016 21:10:18 +0000 (21:10 +0000)]
runtime: fix comment
Fixes #14259
Change-Id: I23fedec0eb85ae28e56bc24539bc864674856130
Reviewed-on: https://go-review.googlesource.com/19318 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Robert Griesemer [Fri, 5 Feb 2016 23:23:47 +0000 (15:23 -0800)]
go/types: make sure constants valid in integer operations are in integer form
The operation where this manifested in a crash was % (only defined on integers).
However, the existing code was sloppy in that it didn't retain the integer form
after a value (e.g., 3.0) was accepted as representable in integer form (3 for
the example). We would have seen a crash in such cases for / as well except
that there was code to fix it for just that case.
Remove the special code for / and fix more generally by retaining the integer
form for all operations if applicable.
Fixes #14229.
Change-Id: I8bef769e6299839fade27c6e8b5ff29ad6521d0d
Reviewed-on: https://go-review.googlesource.com/19300 Reviewed-by: Alan Donovan <adonovan@google.com>
Ian Lance Taylor [Fri, 5 Feb 2016 22:59:46 +0000 (14:59 -0800)]
net/http: deflake TestCloseNotifierPipelined
The test sends two HTTP/1.1 pipelined requests. The first is
completedly by the second, and as such triggers an immediate call to the
CloseNotify channel. The second calls the CloseNotify channel after the
overall connection is closed.
The test was passing fine on gc because the code would enter the select
loop before running the handler, so the send on gotReq would always be
seen first. On gccgo the code would sometimes enter the select loop
after the handler had already finished, meaning that the select could
choose between gotReq and sawClose. If it picked sawClose, it would
never close the overall connection, and the httptest server would hang.
The same hang could be induced with gc by adding a time.Sleep
immediately before the select loop.
Deflake the test by 1) don't close the overall connection until both
requests have been seen; 2) don't exit the loop until both closes have
been seen.
Fixes #14231.
Change-Id: I9d20c309125422ce60ac545f78bcfa337aec1c7d
Reviewed-on: https://go-review.googlesource.com/19281 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Brad Fitzpatrick [Fri, 5 Feb 2016 15:05:35 +0000 (15:05 +0000)]
net/http: fix doc typo
Change-Id: I93201fa4152f2d60b3eedb8d321a152819033121
Reviewed-on: https://go-review.googlesource.com/19270 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Keith Randall [Thu, 4 Feb 2016 21:38:38 +0000 (13:38 -0800)]
runtime: don't expose stack buffer in stringto{byte,rune}slice
When using a stack-allocated buffer for the result, don't
expose the uninitialized portion of it by restricting its
capacity to its length.
The other option is to zero the portion between len and cap.
That seems like more work, but might be worth it if the caller
then appends some stuff to the result. But this close to 1.6,
I'm inclined to do the simplest fix possible.
Fixes #14232
Change-Id: I21c50d3cda02fd2df4d60ba5e2cfe2efe272f333
Reviewed-on: https://go-review.googlesource.com/19231 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
> The stack trace points to it pretty clearly. Done can indeed unblock
> Wait first and then panic. I guess we need to recover after first
> Done as well.
And it looks like TestWaitGroupMisuse2 was already hardened against
this. Do the same in TestWaitGroupMisuse3.
Russ Cox [Tue, 2 Feb 2016 14:19:47 +0000 (09:19 -0500)]
cmd/go: fix rebuild after installation of new Go release
The loading of zversion.go was expecting it to be in
package runtime, but it moved to runtime/internal/sys.
Worse, the load was not checking the error.
Brad Fitzpatrick [Mon, 1 Feb 2016 15:58:34 +0000 (15:58 +0000)]
net/http/httputil: fix spelling of Trailer hop-by-hop header per errata
RFC Errata 4522 (http://www.rfc-editor.org/errata_search.php?eid=4522)
notes that RFC 2616 had a typo in a list of headers that the
httputil.ReverseProxy code copied. Fix the typo in our code.
Austin Clements [Mon, 1 Feb 2016 19:06:51 +0000 (14:06 -0500)]
runtime: start an M when handing off a P when there's GC work
Currently it's possible for the scheduler to deadlock with the right
confluence of locked Gs, assists, and scheduling of background mark
workers. Broadly, this happens because handoffp is stricter than
findrunnable, and if the only work for a P is GC work, handoffp will
put the P into idle, rather than starting an M to execute that P. One
way this can happen is as follows:
0. There is only one user G, which we'll call G 1. There is more than
one P, but they're all idle except the one running G 1.
1. G 1 locks itself to an M using runtime.LockOSThread.
2. GC starts up and enters mark 1.
3. G 1 performs a GC assist, which completes mark 1 without being
fully satisfied. Completing mark 1 causes all background mark
workers to park. And since the assist isn't fully satisfied, it
parks as well, waiting for a background mark worker to satisfy its
remaining assist debt.
4. The assist park enters the scheduler. Since G 1 is locked to the M,
the scheduler releases the P and calls handoffp to hand the P to
another M.
5. handoffp checks the local and global run queues, which are empty,
and sees that there are idle Ps, so rather than start an M, it puts
the P into idle.
At this point, all of the Gs are waiting and all of the Ps are idle.
In particular, none of the GC workers are running, so no mark work
gets done and the assist on the main G is never satisfied, so the
whole process soft locks up.
Fix this by making handoffp start an M if there is GC work. This
reintroduces a key invariant: that in any situation where findrunnable
would return a G to run on a P, handoffp for that P will start an M to
run work on that P.
Fixes #13645.
Tested by running 2,689 iterations of `go tool dist test -no-rebuild
runtime:cpu124` across 10 linux-amd64-noopt VMs with no failures.
Without this change, the failure rate was somewhere around 1%.
Performance change is negligible.
name old time/op new time/op delta
XBenchGarbage-12 2.48ms ± 2% 2.48ms ± 1% -0.24% (p=0.000 n=92+93)
The first in the list above is the main fix that's necessary. The
other are two are in the git history but along for the cmd/bundle
ride. The middle CL is well-tested, small (mostly comments),
non-tricky, and almost never seen (since nobody really uses Trailers).
The final CL is just deleting an unused global variable.
Russ Cox [Fri, 29 Jan 2016 17:18:32 +0000 (12:18 -0500)]
cmd/go: avoid a few symlink-induced errors in internal and vendor checks
This CL expands symlinks only when an error would be reported otherwise.
Since the expansions are only on error paths, anything that worked yesterday
should still work after this CL.
This CL fixes a regression from Go 1.5 in "go run", or else we'd probably
postpone it.
Changing only the error paths is meant as a way to reduce the risk of
making this change so late in the release cycle, but it may actually be
the right strategy for symlinks in general.
Russ Cox [Fri, 29 Jan 2016 15:16:24 +0000 (10:16 -0500)]
cmd/vet: report uncalled functions in Printf %v
Given, say, var f *os.File, a new vet check in CL 14122 diagnoses:
fmt.Printf("%s\n", f.Name)
fmt.Println(f.Name)
but not
fmt.Printf("%v\n", f.Name)
In all three cases the error is that the argument should be f.Name().
Diagnosing Println but not Printf %v seems oddly inconsistent,
so I changed %v to have the check too. In fact, all verbs now have
the check except %p and %T.
Fixes Dave Cheney's confusion when trying to write an example
of the new vet check advertised in the Go 1.6 release notes.
Rahul Chaudhry [Fri, 29 Jan 2016 00:33:35 +0000 (16:33 -0800)]
unsafe: fix typo in documentation of valid Pointer->uintptr->Pointer conversions
Change-Id: Ib669d5241372326a46361ee096570e960b7a957f
Reviewed-on: https://go-review.googlesource.com/19082 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Thu, 28 Jan 2016 06:00:59 +0000 (22:00 -0800)]
runtime: align stack in sigfwd for darwin/386
We might be forwarding to a C signal handler. C code expects the stack
to be aligned. Should fix darwin/386 build: the testcarchive tests were
hanging as the program got an endless series of SIGSEGV signals.
Change-Id: Ia02485d3736a3c40e12259f02d25f842cf8e4d29
Reviewed-on: https://go-review.googlesource.com/19025
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Crawshaw <crawshaw@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Ian Lance Taylor [Wed, 27 Jan 2016 22:07:25 +0000 (14:07 -0800)]
runtime: handle kindString in cgoCheckArg
It's awkward to get a string value in cgoCheckArg, but SWIG testing
revealed that it is possible. The new handling of extra files in the
ptr.go test emulates what SWIG does with an exported function that
returns a string.
Change-Id: I453717f867b8a49499576c28550e7c93053a0cf8
Reviewed-on: https://go-review.googlesource.com/19020
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Dmitry Vyukov [Wed, 27 Jan 2016 18:22:28 +0000 (19:22 +0100)]
runtime/race: run tests with GOMAXPROCS=1
We set GOMAXPROCS=1 to prevent test flakiness.
There are two sources of flakiness:
1. Some tests rely on particular execution order.
If the order is different, race does not happen at all.
2. Ironically, ThreadSanitizer runtime contains a logical race condition
that can lead to false negatives if racy accesses happen literally at the same time.
Tests used to work reliably in the good old days of GOMAXPROCS=1.
So let's set it for now. A more reliable solution is to explicitly annotate tests
with required execution order by means of a special "invisible" synchronization primitive
(that's what is done for C++ ThreadSanitizer tests). This is issue #14119.
This reduces flakes on RaceAsFunc3 test from 60/3000 to 1/3000.
Richard Miller [Wed, 27 Jan 2016 19:10:11 +0000 (19:10 +0000)]
runtime: remove redundant empty function call from Breakpoint on arm
CL 18964 included an extra patch (sorry, my first experience of
git-codereview) which defined the conventional breakpoint instruction
used by Plan 9 on arm, but also introduced a benign but unneeded
call to runtime.emptyfunc. This CL removes the redundant call again.
This completes the series of CLs which add support for Plan 9 on arm.
Russ Cox [Wed, 27 Jan 2016 15:05:18 +0000 (10:05 -0500)]
cmd/go: refine definition of 'standard' import paths to include vendored code
The vendored copy of golang.org/x/net/http/hpack was being treated
as not standard, which in turn was making it not subject to the mtime
exception for rebuilding the standard library in a release, which in turn
was making net/http look out of date.
One fix and three tests:
- Fix the definition of standard.
- Test that everything in $GOROOT/src/ is standard during 'go test cmd/go'.
(In general there can be non-standard things in $GOROOT/src/, but this
test implies that you can do that or you can run 'go test cmd/go',
but not both. That's fine.)
- Test that 'go list std cmd' shows our vendored code.
- Enforce that no standard package can depend on a non-standard one.
Also fix a few error printing nits.
Fixes #13713.
Change-Id: I1f943f1c354174c199e9b52075c11ee44198e81b
Reviewed-on: https://go-review.googlesource.com/18978 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Brad Fitzpatrick [Tue, 26 Jan 2016 19:57:19 +0000 (19:57 +0000)]
net/http: add protections against misuse of ServeFile
Martin Lenord pointed out that bad patterns have emerged in online
examples of how to use ServeFile, where people pass r.URL.Path[1:] to
ServeFile. This is unsafe. Document that it's unsafe, and add some
protections.
Brad Fitzpatrick [Wed, 27 Jan 2016 00:09:50 +0000 (00:09 +0000)]
runtime: add more debug info to flaky TestNumGoroutine
This has been flaking on the new OpenBSD 5.8 builders lately:
https://storage.googleapis.com/go-build-log/808270e7/openbsd-amd64-gce58_61ce2663.log
(as one example)
Austin Clements [Tue, 26 Jan 2016 19:44:58 +0000 (14:44 -0500)]
runtime: make p.gcBgMarkWorker a guintptr
Currently p.gcBgMarkWorker is a *g. Change it to a guintptr. This
eliminates a write barrier during the subtle mark worker parking dance
(which isn't known to be causing problems, but may).
Austin Clements [Tue, 26 Jan 2016 22:26:55 +0000 (17:26 -0500)]
runtime: acquire stack lock in traceEvent
traceEvent records system call events after a G has already entered
_Gsyscall, which means the garbage collector could be installing stack
barriers in the G's stack during the traceEvent. If traceEvent
attempts to capture the user stack during this, it may observe a
inconsistent stack barriers and panic. Fix this by acquiring the stack
lock around the stack walk in traceEvent.
Austin Clements [Wed, 20 Jan 2016 03:45:37 +0000 (22:45 -0500)]
runtime: attach mark workers to P after they park
Currently mark workers attach to their designated Ps before parking,
either during initialization or after performing a phase transition.
However, in both of these cases, it's possible that the mark worker is
running on a different P than the one it attaches to. This is a
problem, because as soon as the worker attaches to a P, that P's
scheduler can execute the worker. If the worker hasn't yet parked on
the P it's actually running on, this means the worker G will be
running in two places at once. The most visible consequence of this is
that once the first instance of the worker does park, it will clear
g.m and the second instance will crash shortly when it tries to use
g.m.
Fix this by moving the attach to the gopark callback. At this point,
the G is genuinely stopped and the callback is running on the system
stack, so it's safe for another P's scheduler to pick up the worker G.
Russ Cox [Tue, 26 Jan 2016 20:26:09 +0000 (15:26 -0500)]
cmd/internal/obj/arm64: adjust literal pool flush for span-dependent jump enlargement
The current code delays the literal pool until the very last moment,
but based on the assumption that span-dependent jumps are as
short as possible. If they need to be enlarged in a later round, that
very last moment may be too late. Flush a little early to prevent that.
Fixes #13579.
Change-Id: I759b5db5c43a977bf2b940872870cbbc436ad141
Reviewed-on: https://go-review.googlesource.com/18972 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Russ Cox <rsc@golang.org>
Adam Langley [Mon, 11 Jan 2016 02:30:27 +0000 (18:30 -0800)]
crypto/rsa: expand on documentation and add some examples.
In some cases the documentation for functions in this package was
lacking from the beginning and, in order cases, the documentation didn't
keep pace as the package grew.
This change somewhat addresses that.
Updates #13711.
Change-Id: I25b2bb1fcd4658c5417671e23cf8e644d08cb9ab
Reviewed-on: https://go-review.googlesource.com/18486 Reviewed-by: Rob Pike <r@golang.org> Reviewed-by: Andrew Gerrand <adg@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Brad Fitzpatrick [Tue, 26 Jan 2016 22:56:08 +0000 (22:56 +0000)]
doc: mention the need for a C compiler for cgo support
Fixes #13954
Change-Id: I4c01e9bb3fb08e8b9fa14d4c59b7ea824ba3f0c9
Reviewed-on: https://go-review.googlesource.com/18937 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Rob Pike <r@golang.org>
Austin Clements [Fri, 15 Jan 2016 19:23:43 +0000 (14:23 -0500)]
runtime/pprof: retry failed tests with longer duration
Currently we run profiling tests for around 200ms in short mode.
However, even on platforms with good profiling, these tests are
inherently flaky, especially on loaded systems like the builders.
To mitigate this, modify the profiling test harness so that if a test
fails in a way that could indicate there just weren't enough samples,
it retries with a longer duration.
This requires some adjustment to the profile checker to distinguish
"fatal" and "retryable" errors. In particular, we no longer consider
it a fatal error to get a profile with zero samples (which we
previously treated as a parse error). We replace this with a retryable
check that the total number of samples is reasonable.