Activity
Yesterday
dmitshur opened a change golang.org/x/build/cmd/relui: update to ltbsky v0.4.22h
Pull in https://github.com/fflewddur/ltbsky/pull/5.dmitshur commented on simd/archsimd: Load*SlicePart panics with "checkptr: converted pointer straddles multiple allocations"3h
CC @golang/compiler.dmitshur commented on cmd/fix: `go fix` inline producing conflicting var names3h
Thanks for the report.
CC @golang/tools-team.dmitshur commented on golang.org/x/build/main.star: remove known issue for linux-arm64_debian13 at tip7h
(1 comment)
This isn't right: the known issue should keep this an optional security trybot. Done in PS 2.dmitshur commented on golang.org/x/build/main.star: remove known issue for linux-arm64_debian13 at tip8h
Thanks.dmitshur commented on cmd/internal/testdir:1_10: Test/codegen/bits.go failures [consistent failure]8h
This consistent failure on ppc64 and loong64 is due to https://go-review.googlesource.com/c/go/+/750200/comments/58cbcdce_f20cfeba.
CC @Jorropo, @golang/ppc64, @golang/loong64.dmitshur commented on crypto: disallow RegisterHash with hash value 08h
(1 comment)
FWIW since its value is actually 20, I understand this means "just use its value by picking a value that's higher than maxHash by some reasonable margin so that it'll be some time before…dmitshur reviewed +2 on cmd/go: specify full path to go command when running go tool covdata9h
Thanks.
Thanks for the clarification.dmitshur commented on crypto: disallow RegisterHash with hash value 09h
(1 comment)
Did something go wrong with your "gofmt-on-save"?dmitshur commented on cmd/go: specify full path to go command when running go tool covdata9h
(1 comment)
Or, maybe the problem being fixed is that once someone with an old toolchain upgrades _to_ 1.25.x or 1.26.x (via go.dev/doc/toolchain), this change (once backported to 1.25.x and 1.26.x)…dmitshur commented on cmd/go: specify full path to go command when running go tool covdata9h
(1 comment)
A question about this; I'm not sure I understood this sequence properly. In the issue comment you described the problem as affecting Go 1.24 and older toolchains upgrading to Go 1.25+. B…dmitshur opened a change golang.org/x/build/main.star: remove known issue for linux-arm64_debian13 at tip10h
Do keep a known issue for Go 1.26 and 1.25 release branches since
it doesn't work there yet (but may work soon).
For golang/go#74985.dmitshur reviewed +2 on golang.org/x/build/main.star: don't run newinliner builder by default in presubmit10h
Thanks.dmitshur reviewed +2 on debug/buildinfo: remove macho/elf DataStart fallback code10h
Thanks.
I think this is right. You can give some more time for another review since Cherry is more familiar with this code, up to you.dmitshur commented on cmd/link: use bfd ld 2.41+ on linux/arm64 instead of gold11h
net/smtp failures are likely #77504, so rebasing should resolve them.dmitshur commented on cmd/link: stop requiring gold on arm64 when GNU ld is fixed11h
@gopherbot Please consider backporting this issue (fixed by [CL 740480](https://go.dev/cl/740480) with a potential followup in [CL 747660](https://go.dev/cl/747660)). The linux/arm64 builder with Deb…dmitshur reviewed +2 on golang.org/x/build/main.star: remove known issue of simd builders11h
Thanks.
Sounds good. (It's possible to follow up in the unlikely case it does need attention.) Thanks for taking release branch into consideration when removing the known issue.dmitshur opened a change cmd/link: [release-branch.go1.25] use bfd ld 2.36+ on linux/arm64 instead of gold12h
The bfd linker has been fixed for a while. In the mean time gold got
deprecated and has stopped receiving new features. Add runtime version
checking and only use gold, if bfd ld 2.35 and lower is det…dmitshur commented on golang.org/x/build: add Debian Trixie builders12h
That problem was tracked in #22040, and [CL 740480](https://go.dev/cl/740480) resolved it. The [linux-arm64_debian13](https://ci.chromium.org/ui/p/golang/builders/ci/gotip-linux-arm64_debian13) is no…dmitshur commented on runtime/race: TestOutput failures13h
A race reported during this test's execution. The test has otherwise been passing without reported races.
CC @golang/runtime.dmitshur commented on internal/godebugs: TestAll failures13h
It looks like the internal/godebugs package tests often come quite close to the 6 minute timeout, e.g., past runs have taken 325.885s and similar. And in this case it went over the time limit, causin…dmitshur commented on cmd/go: test failure due to unexpected telemetry files left behind in temporary directory13h
This looks similar to #55260, but in this case it is some telemetry files being left behind. So far only happened once on a linux-ppc64le builder.
CC @golang/command-line.dmitshur fixed an issue cmd/go: test failure due to unexpected vcstest files left behind in temporary directory13h
There haven't been any matches since 2023, so perhaps this is fixed by various cmd/go test refactoring. There is a new instance in #78397 but it's related to telemetry files rather than vcstest files…dmitshur commented on cmd/go: test failure due to unexpected vcstest files left behind in temporary directory13h
There haven't been any matches since 2023, so perhaps this is fixed by various cmd/go test refactoring. There is a new instance in #78397 but it's related to telemetry files rather than vcstest files…dmitshur commented on golang.org/x/sys/windows: GetNamedSecurityInfo crashes when GetNamedSecurityInfoW succeeds with NULL SECURITY_DESCRIPTOR13h
CC @golang/windows.dmitshur commented on cmd/compile: register allocation differs based on tail loop presence, affects performance13h
CC @golang/compiler.dmitshur commented on syscall: TestPidFD failures13h
This test timed out running for 323 seconds.
From looking at this [test history](https://ci.chromium.org/ui/test/golang/syscall.TestPidFD), it's almost always passing and usually takes something lik…dmitshur commented on cmd/compile: fallthrough prevents the elimination of some bounds checks13h
CC @golang/compiler.dmitshur commented on net/http: how to utilize sendfile() when send request via NewRequestWithContext13h
If this is a question, please see http://go.dev/wiki/Questions for places that are better suited for asking questions. The Go issue tracker is used for bugs, feature requests, optimization issues.
D…dmitshur commented on cmd/compile: channel operations require but do not use a type dictionary13h
CC @golang/compiler.dmitshur commented on cmd/compile: loop induction with `*` and `<<` induction operators13h
CC @golang/compiler.dmitshur commented on cmd/compile: closures in generic functions capture type dictionaries unconditionally13h
CC @golang/compiler.dmitshur commented on archive/zip: provide way to limit number of entries in an archive13h
CC @dsnet, @bradfitz.dmitshur commented on gccgo: internal compiler error in flatten_append on undefined variable in slice of append argument13h
CC @ianlancetaylor, @thanm, @cherrymui.dmitshur commented on gccgo: Compiler Segfault on non-integer argument to make of channel13h
CC @ianlancetaylor, @thanm, @cherrymui.dmitshur commented on cmd/compile: constant-string doubling plus []rune conversion can exhaust memory13h
CC @golang/compiler.dmitshur reviewed +2 on golang.org/x/build/main.star: add linux-amd64 SIMD builders13h
Thanks.dmitshur commented on govulncheck-action: Release current master to have tagged version utilizing the full commit SHAs of transitive actions14h
CC @golang/vulndb, @catzkorn, @lufia.dmitshur commented on os: running "go test os -short" command fails14h
CC @golang/windows, @ianlancetaylor.dmitshur commented on os/user: GroupIds fails when a member of more than 256 groups on macOS14h
CC @golang/darwin, @bradfitz, @kevinburke.This Week
dmitshur commented on golang.org/x/build/cmd/watchflakes: stop applying NeedsInvestigation label1d
This change is deployed now.dmitshur commented on build: build failure on x_vulndb-gotip-linux-amd64-race1d
Something went wrong with the builder infrastructure in this instance. The x/vulndb tests passed, but the task log failed at the end with:
```
RPC failed permanently: rpc error: code = DeadlineExcee…dmitshur closed a change math/big: [release-branch.go1.26] Improve performance of modular exponentiation, particularly for medium-sized inputs.1d
Release branches are used for backport CLs (https://go.dev/wiki/MinorReleases). Closing in favor of CL 751120, which is the same change that targets the main development branch.dmitshur closed a change net/url: [release-branch.go1.26] permit colons in the host subcomponent of non-http/https URLs1d
Replaced with CL 759662.dmitshur commented on net/url: [release-branch.go1.26] permit colons in the host subcomponent of non-http/https URLs1d
This change is ready for review.dmitshur commented on runtime/race: [release-branch.go1.26] apply LLVM zero-initialization fix1d
It seems ready now, so dropping hold.dmitshur commented on golang.org/x/build/main.star: remove known issue for darwin-amd64_151d
Thanks.dmitshur opened a change golang.org/x/build/main.star: remove known issue for darwin-amd64_151d
A builder is added and it's working. More instances may get added later.
Fixes golang/go#77997.dmitshur fixed an issue golang.org/x/build: add darwin/arm64 builders with macOS 26 (macOS Tahoe)1d
dmitshur fixed an issue golang.org/x/build: add darwin/arm64 builders with macOS 26 (macOS Tahoe)1d
The work needed to add darwin/amd64 builders with macOS 26 is somewhat different, so factored that out into its own tracking issue #78388. Retitled this to be about darwin/arm64 only, which is done. …dmitshur opened an issue golang.org/x/build: darwin/amd64 builders with macOS 26 (macOS Tahoe)1d
This is a tracking issue for darwin/amd64 builders with macOS 26 (macOS Tahoe). (We already have darwin/arm64 builders with this macOS version as of issue #76798.)
This is the very last macOS versio…dmitshur reviewed +2 on golang.org/x/build/main.star: add go-private image mirror1d
Thanks.dmitshur commented on runtime/race: [release-branch.go1.26] apply LLVM zero-initialization fix1d
Hold for CL 758160 readiness. (To make tracking easier.)dmitshur opened a change golang.org/x/build/main.star: remove known issue for linux-amd64-longtest-git2.24.01d
The builder has been added and it's working.
For golang/go#26746.
For golang/go#77284.
Fixes golang/go#26653.dmitshur commented on crypto/internal/fips140/drbg: unnecessary linear memory increase on Wasm with Go 1.261d
CC @golang/wasm.dmitshur commented on internal/trace: preallocate pcs array to correct size1d
Thanks.dmitshur commented on net/url: url.Parse in 1.26 breaks mongodb connection string parsing with multiple hosts [1.26 backport]1d
@neild On the main branch, CL 751360 added the mongo-specific implementation and a test, and CL 758900 generalized it for non-http/https and kept the test. The Go 1.26 backport CL 758901 adds the new…dmitshur commented on golang.org/x/build: add a builder with oldest git supported by cmd/go to catch regressions1d
Now that #78323 is resolved, the builder is passing [at tip](https://ci.chromium.org/ui/p/golang/builders/ci/gotip-linux-amd64-longtest-git2.24.0).
It's also passing on [release-branch.go1.26](https…dmitshur commented on cmd/go: [release-branch.go1.25] tests: rename git-min-vers->git-sha2561d
This change is ready for review.dmitshur commented on cmd/go: [release-branch.go1.25] skip git sha256 tests if git < 2.291d
This change is ready for review.dmitshur commented on cmd/go/internal/vcweb/vcstest: TestScripts/git/gitrepo-sha256.txt failures [consistent failure]1d
@gopherbot Please consider this for backport to Go 1.25. (Go 1.26 already has these fixes, so nothing to backport there.) These are test-only fixes needed to make tests pass with the oldest suppor…dmitshur opened a change cmd/go: [release-branch.go1.25] tests: rename git-min-vers->git-sha2562d
Follow-up cleanup on go.dev/cl/698835. Clarify the tests by replacing
git-min-vers with a more-purposeful git-sha256 verb in the cmd tests.dmitshur opened a change cmd/go: [release-branch.go1.25] skip git sha256 tests if git < 2.292d
Fix test building on older Ubuntu LTS releases (that are still
supported). Git SHA256 support was only included in 2.29, which came out
in 2021. Check the output of `git version` and skip these tests…dmitshur reviewed +2 on cmd/dist: try to use jj commit info for devel version2d
This addition seems reasonable since this code path affects Go developers only. Note that it's untested by builders, so it'll be up to whoever uses it locally to maintain. Thanks.
Fixes #nnn? This s…dmitshur commented on golang.org/x/build/main.star: add x/website presubmit builder for all of GOROOT/doc2d
Thanks.dmitshur opened an issue cmd/compile: multiple "internal compiler error" issues2d
There have been a number of "internal compiler error" issues reported recently, each covering specific edge cases that were uncovered. This is an umbrella issue to make it easier to track them.
CC @…dmitshur commented on cmd/go: do not forward GOAUTH-derived credential headers across cross-host redirects2d
From triage, it sounds like the redirect happens by an initially trusted source, and not forwarding GOAUTH may cause it to not work for the target. So there may not be any behavior change to do here.dmitshur commented on cmd/compile: internal compiler error: map elem too big 21474836482d
CC @golang/compiler.dmitshur commented on cmd/link: TestMachOBuildVersion failures2d
From triage, maybe this test doesn't need to run on GOARCH=ppc64le as well.
CC @golang/ppc64.dmitshur commented on cmd/link: peCreateExportFile generates invalid .def file when output name has trailing dot (c-shared on Windows)2d
CC @golang/windows.dmitshur commented on syscall: wrong constant used in `WaitStatus.Stopped()` & `WaitStatus.Continued()` comparisons2d
CC @golang/freebsd, @golang/openbsd.dmitshur commented on crypto/tls: Seg Fault In TLS Handshake in RKE2d
Are you able to reproduce this when using either Go 1.26 or 1.25? In the go env output, it looked like the Go toolchain version was 1.22.dmitshur commented on golang.org/x/website/cmd/golangorg: support absolute paths in testTree2d
Thanks.dmitshur commented on golang.org/x/website/cmd/golangorg: include GOROOT/doc files in TestAll checks2d
Thanks.dmitshur commented on golang.org/x/website/cmd/golangorg: catch instances of bad escaping due to template use2d
Thanks.
The 1.26/1.25 (only) builders failing was [expected](https://go-review.googlesource.com/c/website/+/746741/comments/ae57427c_0909aa7c). The backport CLs have just landed, so everything shoul…dmitshur opened a change cmd/compile: [release-branch.go1.26] skip race detector test failure for unsupported VMA2d
For #77799.
For #78219.dmitshur commented on cmd/compile: panic on invalid generic append with type parameter spread [1.26 backport]2d
Need to confirm whether 1.25 backport is also needed, or if this is 1.26-only per https://github.com/golang/go/issues/34536#issuecomment-572188389.dmitshur commented on cmd/compile: panic on invalid generic append with type parameter spread2d
Does the problem also affect Go 1.25? In that case we should also backport to 1.25 per our policy.dmitshur commented on golang.org/x/build/main.star: remove linux-amd64-spectre known issue2d
It passed at tip (https://ci.chromium.org/b/8686347220472067905) and correctly detected a problem intentionally re-introduced in CL 759121, so everything seems working as intended to me.
Thanks.dmitshur commented on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test2d
Thanks.dmitshur reviewed +1 on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test2d
dmitshur opened a change golang.org/x/sys/cpu: add runtime import in cpu_darwin_arm64_other.go2d
It was missed in CL 737260.
For golang/go#43046.dmitshur commented on golang.org/x/sys/cpu: use IsProcessorFeaturePresent to calculate ARM64 on windows2d
(1 comment)
I agree it would be good to update mkwinsyscall to support this use case, and that it can happen later. In the meantime, what do you think about temporarily replacing the need for manual…dmitshur reviewed +1 on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test2d
dmitshur commented on runtime: memory corruption leading to panic on linux/riscv642d
Thanks for the report. Does this happen only on the linux/riscv64 port specifically, or are you able to reproduce it on other ports too?
CC @golang/riscv64.dmitshur commented on security: fix CVE-2026-338102d
CC @golang/security, @golang/release.dmitshur commented on golang.org/x/build/main.star: set {gc,asm}flags via GO_{GC,TEST_ASM}FLAGS in spectre runmod2d
Thanks.dmitshur commented on cmd/go: add information about GODEBUGs to go help gopath3d
(1 comment)
The issue and this part of the commit message say 'GO111MODULE=auto', but the text being added says 'GO111MODULE=off'. As I understand, standard library tests work with the default moder…dmitshur reviewed +2 on cmd/go: add information about GODEBUGs to go help gopath3d
Thanks.
One too many closing double quotes here.dmitshur commented on golang.org/x/build/cmd/watchflakes: stop applying NeedsInvestigation label3d
(2 comments)
This part of the postNew's documentation seems out of date. Same on line 415.
No need to change it now; just noting that since this ends up being the only place where the constant is n…dmitshur reviewed +2 on golang.org/x/build/cmd/watchflakes: stop applying NeedsInvestigation label3d
Thanks.
I suggest filing some tracking issue for this change, and adding a 'Fixes golang/go#nnn.' line here.
This change will add to the incoming volume of issues that end up in the https://go.dev/…dmitshur reviewed +2 on golang.org/x/build/main.star: remove linux-amd64-spectre known issue3d
Thanks. Note that I rebased and regenerated this change in PS 2.dmitshur commented on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test3d
(1 comment)
gogcflags has "all=" implicitly prepended; removed here in PS 4.dmitshur commented on golang.org/x/build/main.star: remove linux-amd64-spectre known issue3d
Done in CL 758840.dmitshur commented on golang.org/x/build/main.star: set {gc,asm}flags via GO_{GC,TEST_ASM}FLAGS in spectre runmod3d
(1 comment)
I missed that GO_GCFLAGS differs from setting -gcflags directly in that its value gets an implicit "all=" prefix [already prepended](https://cs.opensource.google/go/go/+/master:src/cmd/d…dmitshur reviewed +2 on cmd/go/testdata/vcstest: explicitly set invalid-version's branch to main3d
Glad to see the fix ended up this small. Thanks.
It's probably fine to backport then, if this is all it takes to to unblock the builder on 1.26+1.25 release branches too. I'll check if that's the ca…dmitshur commented on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test3d
(1 comment)
Probably cleaner to add a field to tester instead of a global. It was already not being given any value if cmdtest wasn't run, but seems nicer to make it a clear build error if accidenta…dmitshur commented on cmd/dist: add GO_TEST_ASMFLAGS to set -asmflags during dist test3d
(1 comment)
This struct field and associated code is very easy to add anew when/if we have a need to handle it differently per test, but that's not the case so far, so I'll avoid adding it and makin…dmitshur opened a change golang.org/x/build/main.star: make linux-amd64-longtest-git2.24.0 available in presubmit3d
Being able to optionally request it in presubmit should be helpful when
developing fixes for go.dev/issue/78323.
For golang/go#26653.dmitshur opened a change golang.org/x/build/main.star: set {gc,asm}flags via GO_{GC,TEST_ASM}FLAGS in spectre runmod3d
cmd/dist unsets any GOFLAGS value that might be set in the environment
to make the build and test environment more predictable. It is however
possible to provide gcflags to make.bash via the GO_GCFLA…