Elias Naur [Fri, 1 Feb 2013 16:35:33 +0000 (08:35 -0800)]
cmd/6c, cmd/6g: add flag to support large-model code generation
Added the -pic flag to 6c and 6g to avoid assembler instructions that
cannot use RIP-relative adressing. This is needed to support the -shared mode
in 6l.
See also:
https://golang.org/cl/6926049
https://golang.org/cl/6822078
Russ Cox [Fri, 1 Feb 2013 16:34:41 +0000 (08:34 -0800)]
runtime: cgo-related fixes
* Separate internal and external LockOSThread, for cgo safety.
* Show goroutine that made faulting cgo call.
* Never start a panic due to a signal caused by a cgo call.
Russ Cox [Fri, 1 Feb 2013 16:34:21 +0000 (08:34 -0800)]
cmd/go: clean cgo compiler errors
Cut out temporary cgo file in error message.
Show C.foo instead of _Ctype_foo.
Before:
x.go:20[/var/folders/00/05_b8000h01000cxqpysvccm000n9d/T/go-build242036121/command-line-arguments/_obj/x.cgo1.go:19]: cannot use tv.Usec (type int32) as type _Ctype___darwin_suseconds_t in assignment
After:
x.go:20: cannot use tv.Usec (type int32) as type C.__darwin_suseconds_t in assignment
Russ Cox [Fri, 1 Feb 2013 16:34:08 +0000 (08:34 -0800)]
cmd/cgo: fix line number annotations in generated C code
The old version was using go/ast's CommentGroup.Text method,
but that method drops leading blank lines from the result, so that
if the comment looked like one of
//
// syntax error
import "C"
/*
syntax error
*/
import "C"
then the line numbers for the syntax error would be off by the
number of leading blank lines (1 in each of the above cases).
Russ Cox [Fri, 1 Feb 2013 06:02:20 +0000 (22:02 -0800)]
cmd/dist: redirect acid output to file to separate from errors
If runtime's proc.c does not compile, cmd/dist used to show
the compile errors in a sea of acid output, making them impossible
to find. Change the command invocation to write the acid output
to a file, so that the errors are the only thing shown on failure.
Russ Cox [Thu, 31 Jan 2013 22:11:32 +0000 (14:11 -0800)]
cmd/ld: support for linking with host linker
A step toward a fix for issue 4069.
To allow linking with arbitrary host object files, add a linker mode
that can generate a host object file instead of an executable.
Then the host linker can be invoked to generate the final executable.
This CL adds a new -hostobj flag that instructs the linker to write
a host object file instead of an executable.
That is, this works:
go tool 6g x.go
go tool 6l -hostobj -o x.o x.6
ld -e _rt0_amd64_linux x.o
./a.out
as does:
go tool 8g x.go
go tool 8l -hostld ignored -o x.o x.8
ld -m elf_i386 -e _rt0_386_linux x.o
./a.out
Because 5l was never updated to use the standard relocation scheme,
it will take more work to get this working on ARM.
This is a checkpoint of the basic functionality. It does not work
with cgo yet, and cgo is the main reason for the change.
The command-line interface will likely change too.
The gc linker has other information that needs to be returned to
the caller for use when invoking the host linker besides the single
object file.
Russ Cox [Thu, 31 Jan 2013 21:46:12 +0000 (13:46 -0800)]
io: guarantee err == nil for full reads in ReadFull and ReadAtLeast
This is a backwards compatible API change that fixes broken code.
In Go 1.0, ReadFull(r, buf) could return either len(buf), nil or len(buf), non-nil.
Most code expects only the former, so do that and document the guarantee.
Code that was correct before is still correct.
Code that was incorrect before, by assuming the guarantee, is now correct too.
Adam Langley [Thu, 31 Jan 2013 17:54:37 +0000 (12:54 -0500)]
crypto/x509: test for negative RSA parameters.
Someone found software that generates negative numbers for the RSA
modulus in an X.509 certificate. Our error messages were very poor in
this case so this change improves that.
Update #4728
Return more helpful errors when RSA parameters are negative or zero.
Russ Cox [Thu, 31 Jan 2013 16:06:38 +0000 (08:06 -0800)]
cmd/go: many bug fixes
* Reject import paths of the form cmd/x/y.
* Reject 'go install' of command outside GOPATH
* Clearer error rejecting 'go install' of package outside GOPATH.
* Name temporary binary for first file in 'go run' list or for test.
* Provide a way to pass -ldflags arguments with spaces.
* Pass all Go files (even +build ignored ones) to go fix, go fmt, go vet.
* Reject 'go run foo_test.go'.
* Silence 'exit 1' prints from 'go tool' invocations.
* Make go test -xxxprofile leave binary behind for analysis.
* Reject ~ in GOPATH except on Windows.
* Get a little less confused by symlinks.
* Document that go test x y z runs three test binaries.
* Fix go test -timeout=0.
* Add -tags flag to 'go list'.
* Use pkg/gccgo_$GOOS_$GOARCH for gccgo output.
Rémy Oudompheng [Thu, 31 Jan 2013 07:52:46 +0000 (08:52 +0100)]
cmd/8l: fix misassembling of MOVB involving (AX)(BX*1)
The linker accepts MOVB involving non-byte-addressable
registers, by generating XCHG instructions to AX or BX.
It does not handle the case where nor AX nor BX are available.
Nigel Tao [Thu, 31 Jan 2013 03:12:43 +0000 (14:12 +1100)]
exp/cookiejar: update PublicSuffixList doc comment to match the
examples at http://publicsuffix.org/.
That website previously listed pvt.k12.wy.us, but that was an error,
as confirmed by correspondance with submissions@publicsuffix.org, and
the website was fixed on 2013-01-23.
Marcel van Lohuizen [Wed, 30 Jan 2013 20:19:03 +0000 (21:19 +0100)]
exp/locale/collate/tools/colcmp: fixes some discrepancies between
ICU and collate package: ICU requires strings to be in FCD form.
Not all NFC strings are in this form, leading to incorrect results.
Change to NFD instead.
Rémy Oudompheng [Wed, 30 Jan 2013 20:10:19 +0000 (21:10 +0100)]
cmd/gc: fix export data for aggressive inlining.
Export data was broken after revision 6b602ab487d6
when -l is specified at least 3 times: it makes the compiler
write out func (*T).Method() declarations in export data, which
is not supported.
Also fix the formatting of recover() in export data. It was
not treated like panic() and was rendered as "<node RECOVER>".
Akshat Kumar [Wed, 30 Jan 2013 17:41:16 +0000 (09:41 -0800)]
os: don't hold ForkLock across opens on Plan 9
If os.OpenFile holds ForkLock on files that block opens,
then threads that simultaneously try to do fork-exec will
get hung up (until the open succeeds). Blocked opens are
common enough on Plan 9 that protecting against fd leaks
into fork-execs means not being able to do fork-execs
properly in the general case. Thus, we forgo taking the
lock.
Michael Teichgräber [Wed, 30 Jan 2013 17:25:16 +0000 (09:25 -0800)]
net: SplitHostPort: adjust error message for missing port in IPv6 addresses
An hostport of "[::1]" now results in the same error message
"missing port in address" as the hostport value "127.0.0.1",
so SplitHostPort won't complain about "too many colons
in address" anymore for an IPv6 address missing a port.
Andrey Mirtchovski [Wed, 30 Jan 2013 17:10:32 +0000 (09:10 -0800)]
encoding/json: properly unmarshal empty arrays.
The JSON unmarshaller failed to allocate an array when there
are no values for the input causing the `[]` unmarshalled
to []interface{} to generate []interface{}(nil) rather than
[]interface{}{}. This wasn't caught in the tests because Decode()
works correctly and because jsonBig never generated zero-sized
arrays. The modification to scanner_test.go quickly triggers
the error:
without the change to decoder.go, but with the change to scanner_test.go:
$ go test
--- FAIL: TestUnmarshalMarshal (0.10 seconds)
decode_test.go:446: Marshal jsonBig
scanner_test.go:206: diverge at 70: «03c1OL6$":null},{"[=» vs «03c1OL6$":[]},{"[=^\»
FAIL
exit status 1
FAIL encoding/json 0.266s
Jan Ziak [Wed, 30 Jan 2013 17:01:31 +0000 (09:01 -0800)]
runtime: local allocation in mprof.goc
Binary data in mprof.goc may prevent the garbage collector from freeing
memory blocks. This patch replaces all calls to runtime·mallocgc() with
calls to an allocator private to mprof.goc, thus making the private
memory invisible to the garbage collector. The addrhash variable is
moved outside of the .bss section.
Elias Naur [Wed, 30 Jan 2013 16:46:56 +0000 (08:46 -0800)]
6l/5l: PIC and shared library support for the linkers.
Added the -shared flag to 5l/6l to output a PIC executable with the required
dynamic relocations and RIP-relative addressing in machine code.
Added dummy support to 8l to avoid compilation errors
See also:
https://golang.org/cl/6822078
https://golang.org/cl/7064048
James Gray [Wed, 30 Jan 2013 16:29:33 +0000 (08:29 -0800)]
cmd/cgo: allow for stdcall decorated dynimport names
To allow for stdcall decorated names on Windows, two changes were needed:
1. Change the symbol versioning delimiter '@' in cgo's dynimport output to a '#', and in cmd/ld when it parses dynimports.
2. Remove the "@N" decorator from the first argument of cgo's dynimport output (PE only).
Akshat Kumar [Wed, 30 Jan 2013 15:56:08 +0000 (07:56 -0800)]
include: Plan 9: hide any previous definition of Runemax
Runemax is already defined in libc on 64-bit version of
Plan 9, but is not defined on other versions.
To accommodate, we make sure to rename any previous
instance of Runemax and re-define it subsequently.
Akshat Kumar [Wed, 30 Jan 2013 10:53:56 +0000 (02:53 -0800)]
runtime: add support for panic/recover in Plan 9 note handler
This change also resolves some issues with note handling: we now make
sure that there is enough room at the bottom of every goroutine to
execute the note handler, and the `exitstatus' is no longer a global
entity, which resolves some race conditions.
Alan Donovan [Tue, 29 Jan 2013 15:49:16 +0000 (10:49 -0500)]
exp/ssa: make Parameters values, not addresses.
We explicitly spill all parameters to the frame during initial
SSA construction. (Later passes will remove spills.)
We now properly handle local Allocs escaping via Captures.
Dmitriy Vyukov [Tue, 29 Jan 2013 10:57:11 +0000 (14:57 +0400)]
runtime: dump the full stack of a throwing goroutine
Useful for debugging of runtime bugs.
+ Do not print "stack segment boundary" unless GOTRACEBACK>1.
+ Do not traceback system goroutines unless GOTRACEBACK>1.
Carl Shapiro [Mon, 28 Jan 2013 23:47:25 +0000 (15:47 -0800)]
cmd/ld: avoid a segfault when dumping the symbol table
The dumping routine incorrectly assumed that all incoming
symbols would be non-nil and load through it to retrieve the
symbol name. Instead of using the symbol to retrieve a name,
use the name provided by the caller.
Dave Cheney [Mon, 28 Jan 2013 10:05:25 +0000 (21:05 +1100)]
src: add race.bash
Add race.bash so anyone with suitable hardware can run a race detector build. race.bash can be called from the dashboard builder by passing -cmd="race.bash".
Original source for race.bash is here, http://code.google.com/p/go-wiki/wiki/DashboardBuilders
Ian Lance Taylor [Sun, 27 Jan 2013 02:16:43 +0000 (18:16 -0800)]
runtime: use new CNT_MASK in lfstack
This is for SPARC64, a 64-bit processor that uses all 64-bits
of virtual addresses. The idea is to use the low order 3 bits
to at least get a small ABA counter. That should work since
pointers are aligned. The idea is for SPARC64 to set CNT_MASK
== 7, PTR_BITS == 0, PTR_MASK == 0xffffffffffffff8.
Also add uintptr casts to avoid GCC warnings. The gccgo
runtime code is compiled with GCC, and GCC warns when casting
between a pointer and a type of a different size.
John Graham-Cumming [Fri, 25 Jan 2013 18:20:19 +0000 (10:20 -0800)]
net/http: fix Content-Length/Transfer-Encoding on HEAD requests
net/http currently assumes that the response to a HEAD request
will always have a Content-Length header. This is incorrect.
RFC2616 says: "The HEAD method is identical to GET except that
the server MUST NOT return a message-body in the response. The
metainformation contained in the HTTP headers in response to a
HEAD request SHOULD be identical to the information sent in
response to a GET request. This method can be used for
obtaining metainformation about the entity implied by the
request without transferring the entity-body itself. This
method is often used for testing hypertext links for validity,
accessibility, and recent modification."
This means that three cases are possible: a Content-Length
header, a Transfer-Encoding header or neither. In the wild the
following sites exhibit these behaviours (curl -I):
HEAD on http://www.google.co.uk/ has Transfer-Encoding: chunked
HEAD on http://www.bbc.co.uk/ has Content-Length: 45247
HEAD on http://edition.cnn.com/ has neither header
This patch does not remove the ErrMissingContentLength error
for compatibility reasons, but it is no longer used.
Andrew Gerrand [Thu, 24 Jan 2013 23:06:18 +0000 (10:06 +1100)]
misc/dashboard/builder: synchronize accesses to goroot, always -commit
This prevents the occasional issue when Mercurial screws up the locking
itself, and by moving the locking into this process we can use the
goroot for other things (such as automatically updating the builder
binary).
It also asks all builders to poll for new commits.
Alan Donovan [Thu, 24 Jan 2013 19:22:17 +0000 (14:22 -0500)]
go/types: expose types.IsIdentical, the Type equivalence relation.
This function is absolutely critical for clients such as
exp/ssa, and too complex for clients to duplicate.
As with CL 7200046, gri expressed in the doc below [gophers
only] before going on leave that he intended to expose such a
predicate, though his wording suggests as an interface method
of Type rather than a standalone function. (My preference is
for binary methods to be standalone; see "On Binary Methods",
Kim Bruce, 1995). In any case if he wishes to move it that's
easily accommodated by clients.
Alan Donovan [Thu, 24 Jan 2013 19:21:51 +0000 (14:21 -0500)]
go/types: add String() method to Type interface.
All implementations delegate to typeString.
Though I don't wish to exploit gri's absence to change
his code, this change is pretty low-risk and he assented to it
in the blue ink in the doc below [gophers only].
https://docs.google.com/a/google.com/document/d/1-DQ4fxlMDs9cYtnkKhAAehX6MArjOQyJsRXp-6kiJLA/edit#