[pull] master from golang:master - #163
Merged
Merged
Conversation
This will make it easier to move the Func type into a separate package by removing the dependency from the Func onto the HTMLWriter. The HTMLWriter is just passed around by the ssagen/compile functions. For #80409 Change-Id: I25e2227d197d4b192194ba12f5ac2e086a6a6964 Reviewed-on: https://go-review.googlesource.com/c/go/+/804260 LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Michael Matloob <matloob@google.com>
This change adds a simple helper for running passes in a test without doing a full compile. This will allow splitting ssa.Compile into a separate package, ssacompile, in the next cl. For #80409 Change-Id: I17b7be7cfb0502184c6a3636e52952c06a6a6964 Reviewed-on: https://go-review.googlesource.com/c/go/+/806161 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Michael Matloob <matloob@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
In the x509-limbo path-building tests rather than checking fips140.Version() directly, let's use cryptotest.MustMinimumFIPS140ModuleVersion() with the module version that added ML-DSA support. Cq-Include-Trybots: luci.golang.try:gotip-linux-amd64-longtest Change-Id: I96b5b74b5c9a0a4b69ab2484a5a650aa5dc3a862 Reviewed-on: https://go-review.googlesource.com/c/go/+/808540 LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Auto-Submit: Daniel McCarney <daniel@binaryparadox.net> Reviewed-by: Roland Shoemaker <roland@golang.org> Reviewed-by: Filippo Valsorda <filippo@golang.org>
This commit updates the crypto/internal/cryptotest modules that generate Go code from JSON schemas to use the latest atombender/go-jsonschema release instead of a replace fork. Previously we had this pointed at a PR branch with a bugfix that has since been merged. The updated tooling handles time fields better so after regenerating the x509-limbo schema we can simplify some TestX509Limbo logic. The Wycheproof generated code is 1:1 across tooling updates. Cq-Include-Trybots: luci.golang.try:gotip-linux-amd64-longtest Change-Id: I66d7a0dcb8b69ad8fac79648c4d2a78a0ddd4b97 Reviewed-on: https://go-review.googlesource.com/c/go/+/809460 Reviewed-by: Filippo Valsorda <filippo@golang.org> Reviewed-by: Mark Freeman <markfreeman@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Roland Shoemaker <roland@golang.org>
…pVerify Fixes #80074 Change-Id: I5c619a090293a7668e26699c29b194a56a6a6964 Reviewed-on: https://go-review.googlesource.com/c/go/+/804200 Reviewed-by: Daniel McCarney <daniel@binaryparadox.net> Auto-Submit: Filippo Valsorda <filippo@golang.org> Reviewed-by: Mark Freeman <markfreeman@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
The RPC debug handler now uses slices.SortFunc to sort services and methods. The Len, Less, and Swap methods on serviceArray and methodArray are no longer used. Remove the obsolete sorting methods. This change does not affect the sorting behavior. Change-Id: I8293beb388efb24594d0baf2217fa0d58491112d Reviewed-on: https://go-review.googlesource.com/c/go/+/806660 Reviewed-by: Cherry Mui <cherryyz@google.com> Auto-Submit: Sean Liao <sean@liao.dev> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Sean Liao <sean@liao.dev>
Address.String uses MIME Q encoding for non-ASCII display names unless the name contains a character that cannot appear unencoded in an encoded-word in a phrase. A backslash was missing from this set. mime.QEncoding leaves backslashes unencoded, so Address.String could produce output that ParseAddress could not parse. Select B encoding for display names containing a backslash, and add a round-trip test. Fixes #80592 Change-Id: I53f8fab709bace922914b579c283bb71877c6be7 Reviewed-on: https://go-review.googlesource.com/c/go/+/806681 LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Auto-Submit: Sean Liao <sean@liao.dev> Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Sean Liao <sean@liao.dev> Reviewed-by: Cherry Mui <cherryyz@google.com>
The pkg-config --define-variable flag carries its value inside a single argument. Its validator rejects '-' or '@' anywhere in that value, so it also rejects valid values such as paths containing hyphens, which worked before the pkg-config flag allowlist was added. Allow '-' and '@' after the first character, following the pattern used for --with-path. A leading '-' or '@' is still rejected, since a .pc file may pass the value on to another program, where it could be read as a flag or as a response file. Add direct validator coverage and a script test that exercises the complete cgo build path with a fake pkg-config tool. Fixes #79330 Change-Id: I64d9f5ca9104ee3da73848aa77f6e5d1d684ee1f Reviewed-on: https://go-review.googlesource.com/c/go/+/806860 Reviewed-by: Ian Lance Taylor <iant@golang.org> Auto-Submit: Ian Lance Taylor <iant@golang.org> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
CL 699675 made utf8.DecodeRune and utf8.DecodeRuneInString
inlineable, including their ASCII fast paths. The manual ASCII decoding
fast paths added in CL 793620 are therefore redundant.
Call the UTF-8 decoding functions directly.
On linux/amd64, using benchstat over 10 runs:
goos: linux
goarch: amd64
pkg: bytes
cpu: Intel(R) Core(TM) i7-14700KF
│ old.txt │ new.txt │
│ sec/op │ sec/op vs base │
TrimSpace/SomeNonASCII-20 22.44n ± 1% 22.38n ± 1% ~ (p=0.171 n=10)
TrimSpace/JustNonASCII-20 28.37n ± 1% 28.39n ± 1% ~ (p=0.839 n=10)
geomean 25.23n 25.21n -0.08%
pkg: strings
│ old.txt │ new.txt │
│ sec/op │ sec/op vs base │
TrimSpace/SomeNonASCII-20 25.41n ± 1% 25.38n ± 1% ~ (p=0.305 n=10)
TrimSpace/JustNonASCII-20 31.81n ± 1% 31.82n ± 0% ~ (p=0.813 n=10)
geomean 28.44n 28.42n -0.05%
Change-Id: I317e5bd6d1dcf3a3aed58c8303d1002b71004507
Reviewed-on: https://go-review.googlesource.com/c/go/+/807060
Auto-Submit: Sean Liao <sean@liao.dev>
LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Fix performance regression by restoring original behavior of index function of selecting IndexString vs IndexRabinKarp by introducing guard n>bytealg.MaxLen. At present, the condition else if fails >= 4+i>>4 is reachable when n <= bytealg.MaxLen, which was structurally impossible in 1.26.5, this ends up calling IndexRabinKarp almost all the time leading to slower performance. Regression bisected to the following change https://go-review.googlesource.com/c/go/+/741340 which tried to combine two duplicate loops, adding the guard n > bytealg.MaxLen restores the original behavior while keeping the loops combined. Multiple strings tests improve, none of them regress, pasting performance of a subset of tests that improve the maximum. The change should create similar improvements on other platforms too as it increases the frequency of calling vectorized assembly for index computation. Fixes #80638 goos: linux goarch: amd64 pkg: strings cpu: 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz │ baseline.txt │ new.txt │ │ sec/op │ sec/op vs base │ CountHard1-8 1157.30µ ± 1% 82.65µ ± 6% -92.86% (p=0.000 n=10) CountHard2-8 1159.5µ ± 2% 116.8µ ± 5% -89.93% (p=0.000 n=10) CountHard3-8 1172.3µ ± 2% 373.1µ ± 2% -68.17% (p=0.000 n=10) geomean 1.163m 153.3µ -86.82% Change-Id: Ia9d1b094273751e25bbd2e536ac169363228e7e1 Reviewed-on: https://go-review.googlesource.com/c/go/+/807360 Auto-Submit: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Keith Randall <khr@golang.org> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
…iscv64 Note that ANDN is generated on all riscv64 profiles - the assembler synthesises ANDN in the case the instruction cannot be used directly. Change-Id: I8c6492a24eaa82b30cc631f0ef091fd4c794e774 Reviewed-on: https://go-review.googlesource.com/c/go/+/807900 Reviewed-by: Jorropo <jorropo.pgm@gmail.com> Auto-Submit: Joel Sing <joel@sing.id.au> Reviewed-by: Julian Zhu <jz531210@gmail.com> Reviewed-by: Cherry Mui <cherryyz@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Auto-Submit: Jorropo <jorropo.pgm@gmail.com>
This CL addresses two bugs related to moving goroutine stacks during injected function calls. The runtime injects calls during asynchronous preemption, when handling signals that get turned into panics (sigpanic), and for calls injected by a debugger. If a function triggers a sigpanic, and there is no way to resume execution in the function (such as through a deferred function), then that function's call frame is skipped during stack copying. The reasoning is that its locals, arguments, etc are dead. However, the call frame might contain a frame pointer. We can visit that frame pointer with the execution tracer or block/mutex profilers. If we visit that frame pointer after stack movement, it can point into the old stack. Following it will crash. This CL fixes this by adjusting the frame pointer if there is one. But it still skips the rest of the work, which is unnecessary. For arm64, there is an additonal frame pointer adjustment we're missing. Injected call frames are placed 16 bytes below the stack pointer at the point of call injection. This gap is needed both to avoid clobbering the frame pointer saved below the original function's call frame, and to have space to save the link register so it can be restored when the injected call returns. Normally when function A calls function B, the frame pointer saved below function A's frame is fixed when adjusting function B's frame. But because of the gap in the case of injection, the frame pointer saved by the original function isn't inside any call frame. We need to fix it when visiting the injected call frame. The regression test uses stackPoisonCopy so that the old stack is filled with garbage and frame pointer unwinding will reliably crash. Otherwise we'd only see a crash if something else happens to reuse the old stack space. Fixes #73664 Change-Id: I600d7942521e90852c67e379679b07e96a6a6964 Reviewed-on: https://go-review.googlesource.com/c/go/+/730200 Reviewed-by: Mark Freeman <markfreeman@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Fixes #79225 Change-Id: I51cf9587969af8273a6ebb479f7a0c7f6a6a6964 Reviewed-on: https://go-review.googlesource.com/c/go/+/781844 Reviewed-by: Cherry Mui <cherryyz@google.com> Auto-Submit: Filippo Valsorda <filippo@golang.org> Reviewed-by: Daniel McCarney <daniel@binaryparadox.net> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com>
FormatMediaType lower-cases parameter names when serializing. If the input map contains keys that differ only by case, it can emit duplicate parameter names such as "name=foo; name=bar", which ParseMediaType rejects. Return an empty string in this case, matching FormatMediaType's documented behavior for invalid output. Change-Id: I6d0890bf608da8c3a18af60a36b1825f8dc88ce9 Reviewed-on: https://go-review.googlesource.com/c/go/+/790400 Reviewed-by: shuang cui <imcusg@gmail.com> Reviewed-by: Sean Liao <sean@liao.dev> Reviewed-by: Michael Pratt <mpratt@google.com> Reviewed-by: Mark Freeman <markfreeman@google.com> Auto-Submit: Sean Liao <sean@liao.dev> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
The stream decoder decoded each buffered chunk with a stateless call to Decode, so validity and error offsets depended on how the underlying reader happened to chunk the input. A padded group followed by more input was accepted whenever a chunk boundary fell right after the padding, even though decoding the same input as a whole rejects it, and CorruptInputError offsets were relative to the current chunk rather than the whole stream. Track how much input has been consumed and whether a padded group has been seen: reject any input that follows a padded group, and rebase error offsets so they refer to positions in the whole input stream. Fixes #31626 Change-Id: I2f8c19ac9796fb938430fbcb3df1123583be4a6b GitHub-Last-Rev: a9acf71 GitHub-Pull-Request: #80245 Reviewed-on: https://go-review.googlesource.com/c/go/+/796680 Reviewed-by: Sean Liao <sean@liao.dev> Auto-Submit: Sean Liao <sean@liao.dev> Reviewed-by: Michael Pratt <mpratt@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com>
When using a dictionary, fillWindow was not setting blockStart, which caused a non-compressed block in first position to write the dictionary. This mixup can happen because window stores both the look back window and the block we are about to write. Set blockStart to skip over the dictionary if we have to write the window. Fixes #80538 Change-Id: I6e6784696f29591f2819fc3c9c461d73d3d6335b GitHub-Last-Rev: 469f26b GitHub-Pull-Request: #80539 Reviewed-on: https://go-review.googlesource.com/c/go/+/804680 Auto-Submit: Jorropo <jorropo.pgm@gmail.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Jorropo <jorropo.pgm@gmail.com> Reviewed-by: Mark Freeman <markfreeman@google.com>
ZeroUpper32/48/56Bits kept a hand-maintained per-opcode fact — "this op's result has the upper N bits zero" — in three parallel switches in rewrite.go, far from the op definitions it describes. Every new op had to be added to the right subset of the three lists by hand, which is exactly how MOVBQZX and MOVWQZX ended up missing from the 32-bit list. Declare the fact as a zeroUpperBits attribute on the op definition instead, at the op's strongest valid level; the predicates keep only the value-dependent cases. With the fact at the definition point, an audit of the full op tables extends coverage well beyond the old lists: the flags-to-bool pseudos, the atomics whose result register is written only at load width or by CSET, and value-bounded ops such as CLZ and POPCNT. Codegen validated by diffing -S output of std against the previous commit: on amd64 51 functions improve; on arm64 a net 285 instructions (-848 bytes of text) of re-extensions of atomic results and CSET booleans fold away. Change-Id: I3f2c8b41d19a06f8e027325cc8d64d75e35f9427 Reviewed-on: https://go-review.googlesource.com/c/go/+/806140 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com> Auto-Submit: Keith Randall <khr@golang.org> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Freeman <markfreeman@google.com>
…mory op on amd64
The amd64 backend rewrites a load, a constant modify and a store of the
same memory location into a single read-modify-write instruction, but
only for 32- and 64-bit values: ADDLconstmodify and friends have no 8-
or 16-bit counterpart. So
var tab [256]byte
tab[i&255]++
compiles to a MOVBLZX/INCL/MOVB triple, while the same code over a
[256]uint32 compiles to a single INCL.
Add {ADD,AND,OR,XOR}{B,W}constmodify, their indexed forms, and the
lowering rules that build them. Byte and word arithmetic is done in
32-bit registers, so the rule matches an ADDLconst (etc.) consumed by a
narrow store; the store keeps only the low 8 or 16 bits, which is what
ADDB/ADDW compute. The constant is restricted to values representable
as an 8-bit immediate, so every emitted form encodes as imm8 or as
INC/DEC; that also keeps out the 16-bit immediates that cause
length-changing-prefix stalls.
The store width is unchanged, so the fold never writes bytes the
original store did not.
For the example above:
MOVBLZX AL, AX MOVBLZX AL, AX
LEAQ tab(SB), CX LEAQ tab(SB), CX
MOVBLZX (CX)(AX*1), DX -> INCB (CX)(AX*1)
INCL DX
MOVB DL, (CX)(AX*1)
20 bytes of text become 14.
That is -52 B in runtime.mapassign_fast64, -38 B in mapassign_fast32,
-37 B in mapassign_faststr, -34 B in mapassign_fast64ptr, -32 B in
mapassign and -26 B in internal/runtime/maps.(*Map).Clear. The other
notable cluster is compress/flate, whose [...]uint8 code-length
histogram accounts for 11 of the 12 largest std deltas.
No significant performance difference observed.
386 has the same *Lconstmodify family and the same gap; this change
does not add it.
Updates #10432
Change-Id: I56a0e7cbffac975b253c52326b33fc0eba450e6c
Reviewed-on: https://go-review.googlesource.com/c/go/+/806300
LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Put regular shifts last, so that they can be folded into address calculations. Update #80639 Change-Id: I0a07abbc7a29e1c0c5d6b5c7ed2ae2cffe43957c Reviewed-on: https://go-review.googlesource.com/c/go/+/808200 Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Jorropo <jorropo.pgm@gmail.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
When -buildmode=default is used, the link action ID hashed the literal string "default", so building with -buildmode=default and building with the buildmode it resolves to produced binaries that were identical except for their build IDs. Hash the resolved buildmode (ldBuildmode) instead, so that equivalent builds share a build ID and cached link outputs. Additionally, the build info stamped into binaries normalized -buildmode=default to "exe" without accounting for platforms where the default is PIE (android, ios, darwin, and windows without -race), misreporting the buildmode actually given to the linker. Record "pie" on those platforms instead. Together these make it possible to reproduce a binary built with -buildmode=default from its stamped build info alone. Fixes #63559 Change-Id: If3ac6a6a84fbf20097ff3009e228ce07705d6835 GitHub-Last-Rev: 1ddf5bf GitHub-Pull-Request: #80312 Reviewed-on: https://go-review.googlesource.com/c/go/+/798680 Reviewed-by: Mark Freeman <markfreeman@google.com> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Sean Liao <sean@liao.dev> Reviewed-by: Michael Pratt <mpratt@google.com>
NotExpr.String currently concatenates its leading "!" with the string form of an inner NotExpr. This produces expressions such as "!!linux", which Parse rejects as a double negation. Elide pairs of nested negations while formatting. Account for the effective operator when adding parentheses, so simplifying a negation nested inside an AND or OR expression does not change its meaning. Fixes #80593 Change-Id: Id15885c8b408f8cff143bd994b310523a5cb8452 Reviewed-on: https://go-review.googlesource.com/c/go/+/806682 Reviewed-by: Russ Cox <rsc@golang.org> Reviewed-by: Mark Freeman <markfreeman@google.com> Reviewed-by: Sean Liao <sean@liao.dev> Auto-Submit: Sean Liao <sean@liao.dev> LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
13 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )