Activity
Last Week
dmitshur fixed an issue golang.org/x/build/cmd/coordinator: format %q has arg of wrong type [consistent failure]1d
Fixed by [CL 725080](https://go.dev/cl/725080).dmitshur commented on golang.org/x/build/cmd/coordinator: format %q has arg of wrong type [consistent failure]1d
Fixed by [CL 725080](https://go.dev/cl/725080).dmitshur closed an issue golang.org/x/build/internal/task: unrecognized failures1d
dmitshur closed an issue golang.org/x/build/internal/secret: unrecognized failures1d
dmitshur closed an issue golang.org/x/telemetry/counter/countertest: unrecognized failures [consistent failure]1d
dmitshur fixed an issue golang.org/x/pkgsite/internal/postgres/search: Errorf format %q has arg got of wrong type InputType1d
dmitshur fixed an issue golang.org/x/pkgsite/internal/postgres/search: unrecognized failures1d
Fixed by [CL 725300](https://go.dev/cl/725300).dmitshur commented on golang.org/x/pkgsite/internal/postgres/search: Errorf format %q has arg got of wrong type InputType1d
Fixed by [CL 725300](https://go.dev/cl/725300).dmitshur fixed an issue golang.org/x/text/transform: Errorf format %q has arg nDst of wrong type int1d
dmitshur fixed an issue golang.org/x/text/transform: Errorf format %q has arg nDst of wrong type int1d
Fixed by [CL 725061](https://go.dev/cl/725061).dmitshur commented on golang.org/x/text/transform: Errorf format %q has arg nDst of wrong type int1d
Fixed by [CL 725061](https://go.dev/cl/725061).dmitshur commented on golang.org/x/oscar/internal/gaby: fix %q verb use with wrong type1d
Thanks.dmitshur commented on os: on Unix, Readdirnames skips directory entries with zero inodes1d
@vasi Does this problem apply to Go 1.24 too? If so, we'll need to consider them both per [Go release policy](https://go.dev/issue/34536#issuecomment-572188389).dmitshur commented on os: allow direntries to have zero inodes on Linux1d
@djvasi@gmail.com gopherbot doesn't listen to comments on CLs, only on issues, so you'll need to post this request in issue #76428 itself.
Linking to external discussions is fine, but please be sure…dmitshur opened a change golang.org/x/text/transform: fix %q verb use with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.dmitshur opened a change golang.org/x/telemetry/counter/countertest: fix %q verb use with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.
Fixes golang/go#74262.dmitshur opened a change golang.org/x/pkgsite/internal/postgres: fix %q verb uses with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.dmitshur opened a change golang.org/x/oscar/internal/gaby: fix %q verb use with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.dmitshur commented on golang.org/x/oauth2/endpoints: fix %q verb use with wrong type1d
Thanks.dmitshur opened a change golang.org/x/oauth2/endpoints: fix %q verb use with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.dmitshur opened a change golang.org/x/net/http2: fix %q verb uses with wrong type1d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.
Fixes golang/go#76574.
Fixes golang/go#76599.
Fix…dmitshur commented on cmd/compile: avoid unnecessary interface conversion in bloop1d
This paragraph from there is especially relevant:
> This part of the release cycle is focused on improving the quality of the release, by testing it and fixing bugs that are found. However, every fi…dmitshur commented on cmd/compile: avoid unnecessary interface conversion in bloop1d
Since this is fixing a performance regression in 1.26 itself and we're very early in the freeze, it should be okay unless you think there's a risk of this causing follow up issues and it's better to …dmitshur commented on golang.org/x/build/...: fix %q verb uses with wrong type1d
Thanks.dmitshur commented on golang.org/x/tools/gopls/internal/cache: fix %q verb use with wrong type1d
Thanks.dmitshur reviewed +2 on doc/next: clean up some Go 1.26 release notes2d
Thanks.dmitshur commented on golang.org/x/build/cmd/watchflakes: truncate very large issue body text2d
Thanks.
Agree. This hasn't reached a point where the overhead of a helper function makes up for how much de-duplication it can offer, and such cases tend to become likely upon a 3rd use. (I'm not su…dmitshur opened a change golang.org/x/tools/gopls/internal/cache: fix %q verb use with wrong type2d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.
Fixes golang/go#76547.dmitshur opened a change golang.org/x/build/...: fix %q verb uses with wrong type2d
Caught early by the improved vet check gated behind the 1.26 language
version combined with a tiplang builder that tests with 1.26 language
version.dmitshur closed an issue runtime:cpu2: TestAppendByteCapInLoop failures2d
Merging into #76567.dmitshur closed an issue runtime:cpu1: TestAppendByteCapInLoop failures2d
Merging into #76567.dmitshur opened an issue golang.org/x/build/cmd/watchflakes: default to creating new issues per test (instead of per subtest) and per package (instead of per package:variant)2d
When watchflakes encounters a new failure mode that none of the existing issues in the Test Flakes project match, it creates a new issue for it. When it creates the issue, it creates a pattern for th…dmitshur opened a change golang.org/x/build/cmd/watchflakes: truncate very large issue body text2d
postComment has automatic truncation of very long text for comments,
but GitHub's limit on issue comment length applies to issue bodies too.
This means watchflakes will sometimes error out on issue c…dmitshur commented on golang.org/x/website: typo in blog/generic-interfaces2d
Thanks for reporting. If you'd like, you can contribute a fix. See https://go.dev/doc/contribute.dmitshur commented on doc/next: clean up some Go 1.26 release notes2d
(1 comment)
If you don't mind, also delete this text here. It was accidentally added to doc/initial instead of doc/next during the 1.25 cycle.dmitshur commented on doc/next: clean up some Go 1.26 release notes2d
Thanks.
The word "methods" is already there (on line 1).dmitshur commented on : Tests broken in AIX2d
CC @golang/aix, @ianlancetaylor.dmitshur commented on golang.org/x/tools/gopls/internal/test/marker: Test/*.txt failures2d
Widened the pattern here to catch all Test/*.txt failures, otherwise watchflakes defaults to creating too many individual ones. (If their failure modes end up being different, they can be split up as…dmitshur opened an issue cmd/go/internal/modfetch/codehost: TestReadZip failures2d
```
#!watchflakes
default <- pkg == "cmd/go/internal/modfetch/codehost" && test ~ `^TestReadZip`
```
An issue for watchflakes to track flaky failures of this test.dmitshur commented on net/http: add Transport.NewClientConn3d
(1 comment)
Filed #76480 for it. (watchflakes didn't get to it because it was overwhelmed.)dmitshur opened an issue net/http: TestClientConnReserveAndConsume failures3d
```
#!watchflakes
default <- pkg == "net/http" && test == "TestClientConnReserveAndConsume"
```
Example ([log](https://ci.chromium.org/b/8697100027836472289)):
```
=== RUN TestClientConnReserveAn…dmitshur reviewed +2 on go/build/constraint: use strings.Builder instead of for { str+=str }3d
Thanks.dmitshur commented on net/http: add Transport.NewClientConn3d
(1 comment)
Note that this test seems to be slightly flaky, as visible here:
https://ci.chromium.org/ui/test/golang/net%2Fhttp.TestClientConnReserveAndConsume
(watchflakes doesn't seem to have fil…dmitshur commented on doc/next: document broken freebsd/riscv64 port3d
Thanks.
net/http.TestClientConnReserveAndConsume failure on windows is a recently added test (CL 722223) that seems a bit flaky, watchflakes should file an issue for it soon. Not related to this cha…dmitshur opened a change doc/next: document broken freebsd/riscv64 port3d
Also update comment in cmd/dist's broken map to point to the top-level
umbrella issue.
For #76475.
For #75005.dmitshur opened an issue build: freebsd/riscv64 port is broken3d
The Go porting policy has a section on broken ports, https://go.dev/wiki/PortingPolicy#broken-ports. The GOOS=freebsd GOARCH=riscv64 port was marked broken in August 2025 ([CL 691835](https://go.dev/…dmitshur commented on golang.org/x/build: gotip-freebsd-riscv64 builder is missing3d
@mengzhuo Thanks for working on this. The freebsd/riscv64 builder has passing runs on [some golang.org/x repos](https://ci.chromium.org/ui/p/golang/g/port-freebsd-riscv64/builders), but the main Go r…dmitshur commented on lib/time: update tzdata before release3d
2025b is still the latest available version, so tentatively adding okay-after-rc1 since nothing here blocks the Go 1.26 RC 1 at this time.dmitshur commented on all: announce end of support for old Windows releases3d
Nothing about old versions of Windows to pre-announce in Go 1.26 release notes, moving to the next.dmitshur commented on all: announce end of support for old Linux versions3d
Nothing to pre-announce in Go 1.26 release notes, moving to the next.dmitshur commented on all: announce end of support for old macOS releases3d
[CL 723080](https://go.dev/cl/723080) added a pre-announcement to the Go 1.26 release notes. Moving to the next milestone.dmitshur reviewed +2 on golang.org/x/build/main.star: mark linux-s390x as slow5d
Thanks.Earlier
dmitshur commented on doc: pre-announce removal of 1.23 and earlier crypto GODEBUGs1w
(1 comment)
My understanding is that these removals affect only those users who have opted to modify these GODEBUGs away from their default values in otherwise modern modules (i.e., go directive is …dmitshur reviewed +1 on cmd/go: update default go directive in mod or work init1w
The trybot failures seem unrelated to this change, and each trybot passed just not all at the same time, so it should be safe to bypass in this case.dmitshur reviewed +2 on doc/next: pre-announce end of support for macOS 12 in Go 1.271w
CL 723140 fixes the x/website trybot failure here.dmitshur reviewed +2 on golang.org/x/website/internal/redirect: update expected default redirect to 3071w
Thanks.dmitshur commented on golang.org/x/build/cmd/relui: switch to Gerrit API in UpdateProxyTestRepo1w
Thanks.dmitshur reviewed +2 on doc/next: pre-announce end of support for macOS 12 in Go 1.271w
Thanks.
s/12/13/ here.dmitshur commented on golang.org/x/build/cmd/relui: switch to Gerrit API in UpdateProxyTestRepo1w
(2 comments)
This is a case where it would probably also be fine to just leave it out and let the default behavior apply everywhere. Earlier versions of CreateAutoSubmitChange had a possibility of c…dmitshur reviewed +1 on cmd/go: update default go directive in mod or work init1w
I left a trivial nit, but overall this looks good and I'm not seeing any issues. Thanks for implementing this. Leaving a +1 since I mostly defer to Michael on this review, and he already +2'ed.
This…dmitshur commented on cmd/internal/objabi: remove -V=goexperiment internal special case1w
Thanks.dmitshur commented on golang.org/x/telemetry/internal/telemetry: update test for valid version with GOEXPERIMENT1w
Thanks.dmitshur reviewed +2 on cmd/vendor: update to x/tools@68724af1w
Thanks.dmitshur fixed an issue go/version: does not support the `X:` suffix from `go env GOVERSION`1w
> that telemetry is mixing devel toolchains with "production non-standard" toolchains. The later data set should be more valuable than the former, but (AFAIU) there is no way to diferentiate them.
Y…dmitshur merged a change go/version: use "custom" as an example of a version suffix1w
dmitshur commented on golang.org/x/pkgsite/internal/postgres: avoid locking when upserting the module path1w
There weren't any legacy slowbots requested, so this ends up being an empty set.dmitshur commented on internal/syscall/windows: [release-branch.go1.25] fix ReOpenFile sentinel error value1w
Filed #76360 for it so we don't miss this in the backport review meeting.
Please also add "Fixes #76360." here.dmitshur opened an issue os: on windows RemoveAll removing directories containing read-only files errors with `unlinkat ... Access is denied`, ReOpenFile error handling followup [1.25 backport]1w
This is a follow up for issue #75989, a Go 1.25 backport issue.
The fix for that issue in [CL 715360](https://go.dev/cl/715360) added a `ReOpenFile` syscall to be used by `deleteatFallback`. Its err…dmitshur closed a change cmd/compile: [release-branch.go1.25] during regalloc, fixedreg values are always available1w
Closing since the decision in #75844 is not to backport. If there's new information and we wish to reconsider, it can be reopened then.dmitshur commented on golang.org/x/tools/gopls: in the go repo documentation for functions name `issueXXXXX` should include go.dev/issues link1w
It's also possible to have both (a descriptive test name and a link to a related Go issue number) by moving the Go issue reference into a comment. Something like:
```Go
// Test that <brief descripti…dmitshur reviewed +2 on cmd: update x/tools to 59ff18c2w
Thanks.
(nit) Since this is the main repo, the "golang/go" prefix isn't needed and could be left out. Other than a few extra characters, including it is harmless though.dmitshur commented on golang.org/x/mod/x/mod: apply go fix and go vet2w
Thanks.
It can sometimes be handy to do pure algorithmic changes in patch set 1, and the manual followups in patch set 2, and leave a comment about that. That way the end result code being checked i…dmitshur opened a change golang.org/x/mod/sumdb/note: delete chop2w
CL 720581 replaced almost all instances of it, this gets the last one.
[git-generate]
cd sumdb/note
rf '
ex { import "strings"; chop -> strings.Cut }
rm chop
'dmitshur reviewed +2 on std: fix printf("%q", int) mistakes2w
Thanks.dmitshur reviewed +2 on golang.org/x/mod/x/mod: apply go fix and go vet2w
Thanks. This generally looks good, but I left some suggestions to avoid leaving some of the code in a state where it's doing something in a more verbose/complicated way, making it harder to read. If …dmitshur commented on go/version: use "custom" as an example of a version suffix2w
Thanks.dmitshur opened a change golang.org/x/telemetry/internal/telemetry: update test for valid version with GOEXPERIMENT2w
Add a test case for GOEXPERIMENT toolchain versions after CL 719701.
For golang/go#75953.dmitshur commented on go/version: does not support the `X:` suffix from `go env GOVERSION`2w
@qmuntal On the point about telemetry, I think that is 1) working as intended and 2) there's no change being introduced here.
In the general case, a toolchain built with an experiment may differ in …dmitshur opened a change go/version: use "custom" as an example of a version suffix2w
The suffix in a non-standard toolchain version can be any string. Show
more of a middle ground example of a non-standard version suffix,
aligning it with the example used at https://go.dev/doc/toolch…dmitshur commented on go/version: does not support the `X:` suffix from `go env GOVERSION`2w
The suffix after the dash of a non-standard toolchain can be any string, and at https://go.dev/doc/toolchain#name the example shown uses "custom" rather than "bigcorp". Perhaps it'd be more clear if …dmitshur opened a change cmd/internal/objabi: remove -V=goexperiment internal special case2w
This special case was added in CL 310171 for test/run.go use, as the
comment still says, but run.go (cmd/internal/testdir/testdir_test.go
by now) stopped using this in CL 310732. There don't seem to …dmitshur reviewed +2 on cmd/go: add test checking version with experiment is valid2w
Thanks.
I suspect that it will be possible to cause this test to fail by doing something like:
```
cat "bad" > VERSION
cd src && ./make.bash
```
That is, since it's possible to set a custom versio…dmitshur commented on cmd/link/internal/ld: make runtime.buildVersion with experiments valid2w
(1 comment)
If you want, consider modifying this test case to also check that version.IsValid(runtime.Version()) isn't unexpectedly false.
It's probably not going to catch every instance of regress…dmitshur reviewed +2 on cmd/link/internal/ld: make runtime.buildVersion with experiments valid2w
Thanks.
That works, and has the advantage of being less likely to run into a false positive match.dmitshur commented on runtime/secret: implement new secret package2w
A new builder is typically appropriate when the goal is to run all of the Go tests under a different mode. For example, if the GOEXPERIMENT modifies runtime behavior.
If the change of behavior here …dmitshur reviewed +2 on golang.org/x/tools/internal/analysis/analyzerutil: fix FileUsesGoVersion2w
Thanks.
You added a Auto-Submit+1 vote to patch set 2, which had no effect because patch set 3 was latest, and that vote doesn't transfer between patch sets. Re-adding it to PS3 which I understand w…dmitshur reviewed +2 on golang.org/x/text/...: eliminate vet diagnostics2w
Thanks.
Yes, that seems like a false positive in that vet check, especially since the package documentation at https://pkg.go.dev/testing#hdr-Benchmarks shows examples of for b.Loop() loops, and non…dmitshur reviewed +2 on golang.org/x/tools/internal/analysis/analyzerutil: fix FileUsesGoVersion2w
Thanks.
(nit) What's the motivation to include an underscore prefix in the result bool name?
I don't quite see how this comment is connected to the code, `stdlib.IsBootstrapPackage(pkgpath)` as I i…dmitshur reviewed +2 on golang.org/x/build/main.star: invert greenteagc and sizespecializedmalloc builders2w
Thanks.dmitshur commented on golang.org/x/tools/go/analysis/passes/modernize: respect bootstrap toolchain version2w
(1 comment)
@adonovan@google.com Would it make sense to add a special case here to not apply the lower BootstrapVersion to package tests?
The packages used during bootstrap do need to build and wor…dmitshur reviewed +2 on std: go fix -any std cmd2w
Thanks.
None of these tests were sensitive to whether interface{} or any is used.
This variable name becomes more indirect than it was before, but it seems fine.dmitshur commented on golang.org/x/website/internal/dl: unpredictably applied HTTP transport compression for .tar.gz resources results in negligible bandwidth savings while confusing contemporary versions of Firefox2w
@luismferreira You should confirm that it's not your environment forcing the go1.25.0 toolchain to [be selected](https://go.dev/doc/toolchain#select). Try re-running `go version` with `GODEBUG=toolch…dmitshur reviewed +2 on golang.org/x/text/...: eliminate vet diagnostics2w
Thanks.
Would it be better to resolve the unused result by modernizing this benchmark with the `testing.B.Loop` API?
```
func BenchmarkString(b *testing.B) {
for b.Loop() {
USD.String()
}
}
```…dmitshur reviewed +2 on golang.org/x/perf/...: eliminate vet diagnostics2w
Thanks.
I'll note that compared to other things vet reports, it's not clear to me that report is in this context is very useful. The parse.SyntaxError type is defined in an internal package relative…dmitshur reviewed +2 on golang.org/x/image/...: eliminate vet diagnostics2w
Thanks.
I see. Well, if we're going to rely on being able to generate these files, the generator program probably needs to be added to this repo where it can be modified more easily.
But this packa…dmitshur commented on golang.org/x/image/...: eliminate vet diagnostics2w
(1 comment)
These two files are generated. Won't the change be undone by next 'go generate' invocation? Or is the generator updated already?dmitshur reviewed +2 on golang.org/x/build/...: eliminate vet diagnostics2w
Thanks.dmitshur commented on golang.org/x/website: sha256sum for go1.25.4.linux-amd64.tar.gz download does not match3w
9fa5ffeda4170de60f67f3aa0f824e426421ba724c21e133c1e35d6159ca1bec is the right sha256 checksum for the go1.25.4.linux-amd64.tar.gz file. It can be reproduced with distpack by building Go 1.25.4 from s…dmitshur closed a change path/filepath: [release-branch.go1.24] return cleaned path from Rel3w
Thanks for preparing the CL. Closing since the decision in #75817 is not to backport.dmitshur closed a change path/filepath: [release-branch.go1.25] return cleaned path from Rel3w
Thanks for preparing the CL. Closing since the decision in #75818 is not to backport.