Repository navigation
Conversation
Go 1.27 ships Unicode 17 IDNA tables, which reject an invalid UTF-8 label. \x94.EXAMPLE.COM no longer becomes xn--zn7c.example.com. normalizeHTTPHost already returns that error. The test follows the toolchain the tables were built for. Fixes docker#4138. Signed-off-by: Sasha Mitchell <sash.t.mitchell@gmail.com>
|
I'm going to close this PR because the activity from this account raises concerns about automated PR farming. At the time of writing, GitHub shows 1,092 authored PRs, including 20 opened today (October 12). The latest ten were opened across ten different repositories in under five hours, covering unrelated changes to cryptography, terminal rendering, caching, XPath, and PDF generation. Examples include ProtonMail/go-crypto#338, charmbracelet/lipgloss#759, and antchfx/xpath#158. That doesn't prove the account is a bot, but the volume and pattern are concerning enough that I'm not comfortable merging this contribution. I'll investigate #4138 independently and follow up with a fix under my own account. If I've misread the situation, please reach out and explain how you're personally reviewing and validating these contributions. I'm open to correcting my judgment. |
|
Hey Crazy Max, I left a comment on other PR, I'm real but your right in I'm directing surplus compute with my system to OSS projects, so there is sudden spikes in PRs, it's not to be malicious, so my apologies and I can stop directing your way if preferred In work I have multiple Cursor accounts and whilst using with API quota, I end up with surplus in Grok 4.7 quota, if not used it's a waste imo, we have a codered security system built for internal company work so i just point it to interesting OSS projects, I am here an present and managing over it and checking, but my agent is also quite aggressive in working through PRs and our orchestrator feedback loops Hope this helps on transparency, and if wish to merge yourself or give any feedback (if not helpful or anything i can point towards) that's also no problem too, thanks |
2 participants
Go 1.27 builds
golang.org/x/net/idnawith the Unicode 17 tables. Those tables reject invalid UTF-8. The Unicode 15 tables encoded\x94as U+FFFD, soTestNormalizeHTTPHostexpectedxn--zn7c.example.com. On Go 1.27,normalizeHTTPHostreturnsidna: invalid labelwrapped asinvalid http host.The case now lives in two tests with the same build tags as the IDNA tables. Go 1.26 still expects the punycode. Go 1.27 expects the error.
Before the change,
GOTOOLCHAIN=go1.27.2 go test -mod=vendor -count=1 -run 'TestNormalizeHTTPHost$' ./policyfailed on the\x94.EXAMPLE.COMcase. After it, that run passes, andgo test -mod=vendor -count=1 ./policypasses on Go 1.26.5.Fixes #4138.