Change-Id: I86530a4fd240f3e056e3277355d2965f6863b852
Reviewed-on: https://go-review.googlesource.com/8951Reviewed-by: Rob Pike <r@golang.org>

It is not needed right now, but it will allow more flexibility in
the future.

Fixes #10472

Change-Id: I2eaea70abeca5ed10f89b0b2dfdabdac376a0a41
Reviewed-on: https://go-review.googlesource.com/8964Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

The ptrto field of the type data cannot be relied on when dynamic linking: a
type T may be defined in a module that makes no use of pointers to that type,
but another module can contain a package that imports the first one and does use
*T pointers.  The second module will end up defining type data for *T and a
type.*T symbol pointing at it. It's important that calling .PtrTo() on the
refect.Type for T returns this type data and not some synthesized object, so we
need reflect to be able to find it!

Fortunately, the reflect package already has a mechanism for doing this sort of
thing: ChanOf/MapOf/etc look for pre-existing type data by name.  So this change
just extends PtrTo() to consult this too, and changes the compiler to include
pointer types in the data consulted when compiling for dynamic linking.

Change-Id: I3773c066fd0679a62e9fc52a84bf64f1d67662b7
Reviewed-on: https://go-review.googlesource.com/8232Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

This is a follow-up to CL 8910.
This is the version that I have tested and which works
when appID and teamID are not the same (which they appear
to be for the builder).

I am unsure how I submitted it with the wrong code.

Change-Id: I186e34e91953d082b507390c1cd2042e5419c4c9
Reviewed-on: https://go-review.googlesource.com/8943Reviewed-by: David Crawshaw <crawshaw@golang.org>

7g and 9g disagree with componentgen
about what type len and cap have.

This results in an etype mismatch,
which inhibits registerization.

Fixing this results in 7406 more registerizations
while building the stdlib.
There are still 1512 missed opportunities.

This should improve the performance benefit
to 7g of enabling componentgen (CL 8636).

This CL reduces the size of godoc by 203k (-1.177%).

This was discovered by using the diagnostics
added in CL 8732 and running:

GOARCH=arm64 GOOS=linux go build -gcflags="-d registerization" std

See CL 91850043 for similar earlier fixes for 6g and 8g.

Change-Id: I57f478228a000ad7529d4136bad94a51343c4daa
Reviewed-on: https://go-review.googlesource.com/8733Reviewed-by: Dave Cheney <dave@cheney.net>
Reviewed-by: Keith Randall <khr@golang.org>

We can use CBNZ instruction and make it one instruction shorter.

Saves 66kB in godoc.

Change-Id: Ie71fe7cf31e7f73644ee926f4f9624c009c3eb1a
Reviewed-on: https://go-review.googlesource.com/8634Reviewed-by: Minux Ma <minux@golang.org>

Optimize the sequence:
	MOV $imm, Rt
	ADD Rt, Rs, Rd

into:
	ADD $imm, Rs, Rd

Saves 66k in godoc.

Change-Id: I27b4aaa0ec80a59472fe2e5816efdf3db9c901ee
Reviewed-on: https://go-review.googlesource.com/8632Reviewed-by: Minux Ma <minux@golang.org>

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f03
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8782Reviewed-by: David Crawshaw <crawshaw@golang.org>

Because there is no libgcc.

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f07
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8786Reviewed-by: David Crawshaw <crawshaw@golang.org>

Change-Id: I402629b154ae36c879a8d5088cb0f71b075f31dc
Reviewed-on: https://go-review.googlesource.com/8958Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

Fixes #10116.

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f05
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8784Reviewed-by: David Crawshaw <crawshaw@golang.org>

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f04
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8783Reviewed-by: David Crawshaw <crawshaw@golang.org>

We no longer need the EXC_BAD_ACCESS watcher as runtime/cgo contains
a mach exception handler that catches it. And now lldb only
intermittently reports process connection and exiting, so instead
just look for the PASS from Go.

Change-Id: I403266558f5a900e0b87ec1019d9baec88148d23
Reviewed-on: https://go-review.googlesource.com/8957Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

If the requested coding bit sizes don't result in a full binary tree,
then reject the input as invalid.

Exception: We still need to allow degenerate Huffman codings with a
single 1-bit code to be compatible with zlib and files compressed with
Go's compress/flate package.

Update #10426.

Change-Id: I171b98d12e65b4deb9f4031cd802407ebb5e266c
Reviewed-on: https://go-review.googlesource.com/8922Reviewed-by: Nigel Tao <nigeltao@golang.org>

Change-Id: I6089cebf7de73ce9c4e53ad3e8ef2673d970bda2
Reviewed-on: https://go-review.googlesource.com/8908Reviewed-by: Rob Pike <r@golang.org>

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f02
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8781Reviewed-by: David Crawshaw <crawshaw@golang.org>

ld64 cannot handle BR26 reloc with non-zero addend. It incorrectly
thinks that non-zero addend for BR26 means the code is not PIC, but
those BR26 relocs should be fully resolved at link time.

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f01
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8780Reviewed-by: David Crawshaw <crawshaw@golang.org>

To support "SVC $0x80", which is needed for darwin/arm64.

Change-Id: I3b3f80791a1db4c2b7318f81a115972cd2237f00
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/8769Reviewed-by: David Crawshaw <crawshaw@golang.org>

This code's test coverage is ad hoc at best, and it's easy to make
changes that accidentally regress invariants.  This CL adds a "sanity"
constant that can be changed to "true" during development to add extra
runtime checking that the Huffman decoder tables are sane.

Change-Id: I0d0ca53ad7c9566be18046d9b255e1a30059f28b
Reviewed-on: https://go-review.googlesource.com/8974Reviewed-by: Nigel Tao <nigeltao@golang.org>

When decoding Huffman codes, if an invalid bit sequence is discovered,
reject the input instead of treating it as a 0-length code.

Fixes #10426.

Change-Id: Ie2f1a3a718afd7c6bee73a67480d4b84936c21c9
Reviewed-on: https://go-review.googlesource.com/8893
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Nigel Tao <nigeltao@golang.org>

This also involves adding functions to typelinks along with a minor
change to ensure they are sorted correctly.

Change-Id: I054a79b6498a634cbccce17579f52c299733c2cf
Reviewed-on: https://go-review.googlesource.com/1996Reviewed-by: Ian Lance Taylor <iant@golang.org>

Change-Id: I211c0d33dc292c6a703d788f6d4d286107bcb6b0
Reviewed-on: https://go-review.googlesource.com/8906Reviewed-by: Rob Pike <r@golang.org>

Under some dial tests that require external network connectivity, we
must prevent application traffic but must not interfere with control
plane traffic such as DNS message exchange. But test helper function
disableSocketConnect prevents both application and control plane traffic
unconditionally and makes some dial tests with -ipv6 fail when
CGO_ENABLED=0.

This change makes disableSocketConnect take a look at not only address
family but socket type for fixing some dial tests with -ipv6 when
CGO_ENBALED=0.

Change-Id: I32241d9592d31483424bb5e69cb4d56f3fc20312
Reviewed-on: https://go-review.googlesource.com/8743Reviewed-by: Ian Lance Taylor <iant@golang.org>

Change-Id: I09e84161d106960a69972f5fc845a1e40c28e58f
Reviewed-on: https://go-review.googlesource.com/8331
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

You can now do 'go install -buildmode=shared std' and get yourself
a nice (33 meg) libstd.so (which is not useful until there is -linkshared
support as well, of course).

Change-Id: Ie9b7e7f72abc7d369a6e3ecc98903a9d197bd6e6
Reviewed-on: https://go-review.googlesource.com/8300
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

Move the single file from internal/syscall to internal/syscall/unix,
to match the golang.org/x/sys layout.

Change-Id: I2fb2832b4cb22efc7666bd276f5401ac3e73dc40
Reviewed-on: https://go-review.googlesource.com/8972Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Rob Pike <r@golang.org>

Change-Id: I9ceceb29291ea9f5d7b675dfabd665c5e3618471
Reviewed-on: https://go-review.googlesource.com/8955Reviewed-by: Ian Lance Taylor <iant@golang.org>

Fixes #10449

Change-Id: I1dc2d0213e6a46f3609222d5460c1a54081e2471
Reviewed-on: https://go-review.googlesource.com/8931Reviewed-by: Ian Lance Taylor <iant@golang.org>

This is $GOPATH/pkg/linux_amd64 or similar.  cmd/go already had a grotty calculation
of this and I need to add another one for -buildmode=shared.

Change-Id: Ied28c9b7cce671da8d45920e124a3e0c2501258a
Reviewed-on: https://go-review.googlesource.com/8930Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

Including having -r "" preventing rpath from being set at all.

Change-Id: Ib40d7bf93a6e9ef21985c4a05b5703e4fbd1cd1b
Reviewed-on: https://go-review.googlesource.com/8806Reviewed-by: Ian Lance Taylor <iant@golang.org>

It is faster to execute

	MOVQ AX,(DI)
	MOVQ AX,8(DI)
	MOVQ AX,16(DI)
	MOVQ AX,24(DI)
	ADDQ $32,DI

than

	STOSQ
	STOSQ
	STOSQ
	STOSQ

However, in order to be able to jump into
the middle of a block of MOVQs, the call
site needs to pre-adjust DI.

If we're clearing a small area, the cost
of that DI pre-adjustment isn't repaid.

This CL switches the DUFFZERO implementation
to use a hybrid strategy, in which small
clears use STOSQ as before, but large clears
use mostly MOVQ/ADDQ blocks.

benchmark                 old ns/op     new ns/op     delta
BenchmarkClearFat8        0.55          0.55          +0.00%
BenchmarkClearFat12       0.82          0.83          +1.22%
BenchmarkClearFat16       0.55          0.55          +0.00%
BenchmarkClearFat24       0.82          0.82          +0.00%
BenchmarkClearFat32       2.20          1.94          -11.82%
BenchmarkClearFat40       1.92          1.66          -13.54%
BenchmarkClearFat48       2.21          1.93          -12.67%
BenchmarkClearFat56       3.03          2.20          -27.39%
BenchmarkClearFat64       3.26          2.48          -23.93%
BenchmarkClearFat72       3.57          2.76          -22.69%
BenchmarkClearFat80       3.83          3.05          -20.37%
BenchmarkClearFat88       4.14          3.30          -20.29%
BenchmarkClearFat128      5.54          4.69          -15.34%
BenchmarkClearFat256      9.95          9.09          -8.64%
BenchmarkClearFat512      18.7          17.9          -4.28%
BenchmarkClearFat1024     36.2          35.4          -2.21%

Change-Id: Ic786406d9b3cab68d5a231688f9e66fcd1bd7103
Reviewed-on: https://go-review.googlesource.com/2585Reviewed-by: Keith Randall <khr@golang.org>

Val is used to hold constant values.
Reg was the odd duck out.

Generated using eg.

No functional changes. Passes toolstash -cmp.

Change-Id: Ic1de769a1f92bb02e09a4428d998b716f307e2f6
Reviewed-on: https://go-review.googlesource.com/8912Reviewed-by: Russ Cox <rsc@golang.org>

By removing type slice, renaming type sliceStruct to type slice and
whacking until it compiles.

Has a pleasing net reduction of conversions.

Fixes #10188

Change-Id: I77202b8df637185b632fd7875a1fdd8d52c7a83c
Reviewed-on: https://go-review.googlesource.com/8770Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

We forgot to add the !gccgo tag to cpuid_386.s.

Change-Id: I2de2ed92ac9686c9365cb37cd29121fa98c2bf37
Reviewed-on: https://go-review.googlesource.com/8960Reviewed-by: Dave Cheney <dave@cheney.net>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

Change-Id: Ia0d7c2f08f8f124bcd62c7f9744227812d68903d
Reviewed-on: https://go-review.googlesource.com/8941Reviewed-by: Rob Pike <r@golang.org>

When a reflect.Value is passed to Printf (etc.), fmt called the
String method, which does not disclose its contents. To get the
contents, one could call Value.Interface(), but that is illegal
if the Value is not exported or otherwise forbidden.

This CL improves the situation with a trivial change to the
fmt package: when we see a reflect.Value as an argument,
we treat it exactly as we treat a reflect.Value we make inside
the package. This means that we always print the
contents of the Value as if _that_ was the argument to Printf.

This is arguably a breaking change but I think it is a genuine
improvement and no greater a break than many other tweaks
we have made to formatted output from this package.

Fixes #8965.

Change-Id: Ifc2a4ce3c1134ad5160e101d2196c22f1542faab
Reviewed-on: https://go-review.googlesource.com/8731Reviewed-by: roger peppe <rogpeppe@gmail.com>
Reviewed-by: Russ Cox <rsc@golang.org>

This trivial addition to the io package makes it easy to control the
buffer size and allocation properties of io.Copy.

Change-Id: Ica1a6bd015e429d4e655bc0c6f66cea21c454acf
Reviewed-on: https://go-review.googlesource.com/8730Reviewed-by: Russ Cox <rsc@golang.org>

Fixes #9855

Use an architectural zero register as the source for zeroing, if available.

Change-Id: Ie5b4ba4e3d356c6f892bfd1cebd14d5152bdeeb0
Reviewed-on: https://go-review.googlesource.com/8722Reviewed-by: Keith Randall <khr@golang.org>

Change-Id: I63053c45081f47ba736720a78d85c31c13767891
Reviewed-on: https://go-review.googlesource.com/8642Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

Change-Id: Icaac9a48a3b9a3c5542235162e21ab8303592965
Reviewed-on: https://go-review.googlesource.com/8641Reviewed-by: Ian Lance Taylor <iant@golang.org>