Skip to content

N
net
Switch branch/tag
  1. 24 Mar, 2016 1 commit
  3. 22 Mar, 2016 2 commits
  5. 21 Mar, 2016 1 commit
  7. 20 Mar, 2016 1 commit
  9. 15 Mar, 2016 1 commit
  11. 10 Mar, 2016 1 commit
  13. 09 Mar, 2016 2 commits
  15. 06 Mar, 2016 1 commit
  17. 05 Mar, 2016 1 commit
    • Jacob Hoffman-Andrews's avatar
      publicsuffix: Make gen.go faster. · 08f168e5
      Jacob Hoffman-Andrews authored 
      Previously this took 20 minutes, now it takes 2-4 seconds.

Changes:

 - Iterate over prefix length, from longest to shortest.
 - Build a map of prefixes and reuse it to avoid one loop.
 - When removing simple substrings, sort by length and only consider strings
   long enough to be worth checking. Saves 600ms out of 800ms.
 - Removed the option to generate uncrushed table.
 - Fix an unhandled error in n.walk(w, assignIndexes)

Change-Id: I321d2c2bd18f4918479500f3c61d5e59ee173f3d
Reviewed-on: https://go-review.googlesource.com/20029Reviewed-by: 's avatarNigel Tao <nigeltao@golang.org>
      08f168e5
  19. 04 Mar, 2016 1 commit
  21. 25 Feb, 2016 2 commits
  23. 23 Feb, 2016 3 commits
    • Dmitri Shuralyov's avatar
      http2/h2i: Handle invalid usage more idiomatically. · 4599ae79
      Dmitri Shuralyov authored 
      I believe it is more idiomatic to use exit code 2 for flag parsing
errors and invalid usage cases.

It's also not needed to do os.Exit inside usage since flag handling
code does it anyway; usage should only print usage text.

Change-Id: I0fe2047e3fd01489d32dfb8fde49ce7829439687
Reviewed-on: https://go-review.googlesource.com/19774Reviewed-by: 's avatarBlake Mizerany <blake.mizerany@gmail.com>
      4599ae79
    • Mikio Hara's avatar
      internal/iana: update protocol numbers · 0899459b
      Mikio Hara authored 
      Change-Id: I88b1fac3419f3e655a11cb22050fcd9a9a6c6616
Reviewed-on: https://go-review.googlesource.com/19786Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      0899459b
    • Brad Fitzpatrick's avatar
      http2: move merging of HEADERS and CONTINUATION into Framer · 9e1fb3c1
      Brad Fitzpatrick authored 
      HEADERS and CONTINUATION frames are special in that they must appear
contiguous on the wire and there are lots of annoying details to
verify while working through its state machine, including the handling
of hpack header list size limits and DoS vectors.

We now have three implementations of this merging (Server, Transport,
and grpc), and grpc's is not complete. The Transport's was also
partially incomplete.

Move this up to the Framer (opt-in, for compatibility) and remove the
support from the Server and Transport. I can fix grpc later to use
this.

Recommended reviewing order:

* hpack.go exports the HeaderField.Size method and adds an IsPseudo
  method.

* errors.go adds some new unexported error types, for testing.

* frame.go adds the new type MetaHeadersFrame.

* frame.go adds new fields on Framer for controlling how ReadFrame
  behaves

* frame.go Framer.ReadFrame now calls the new Framer.readMetaFrame
  method

* frame_test.go adds a bunch of tests. these are largely redundant
  with the existing tests which were in server and transport
  before. They really belong with frame_test.go, but I also don't want
  to delete tests in a CL like this. I probably won't remove them
  later either.

* server.go and transport.go can be reviewed in either order at this
  point. Both are the fun part of this change: deleting lots of hairy
  state machine code (which was redundant in at least 6 ways: server
  headers, server trailers, client headers, client trailers, grpc
  headers, grpc trailers...). Both server and transport.go have the
  general following form:

  - set Framer.ReadMetaHeaders
  - stop handling *HeadersFrame and *ContinuationFrame; handle
    *MetaHeadersFrame instead.
  - delete all the state machine + hpack parsing callback hell

The diffstat numbers look like a wash once you exclude the new tests,
but this pays for itself by far when you consider the grpc savings as
well, and the increased simplicity.

Change-Id: If348cf585165b528b7d3ab2e5f86b49a03fbb0d2
Reviewed-on: https://go-review.googlesource.com/19726Reviewed-by: 's avatarBlake Mizerany <blake.mizerany@gmail.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
      9e1fb3c1
  25. 17 Feb, 2016 3 commits
  27. 13 Feb, 2016 1 commit
    • Brad Fitzpatrick's avatar
      http2: fix crash in Transport on double Read of invalid gzip Response.Body · cbbbe2bc
      Brad Fitzpatrick authored 
      This old code was buggy:

type gzipReader struct {
        body io.ReadCloser // underlying Response.Body
        zr   io.Reader     // lazily-initialized gzip reader
}

func (gz *gzipReader) Read(p []byte) (n int, err error) {
        if gz.zr == nil {
                gz.zr, err = gzip.NewReader(gz.body)
                if err != nil {
                        return 0, err
                }
        }
        return gz.zr.Read(p)
}

If a Read on a gzipped Response.Body (of type *http2.gzipReader)
resulted in gzip.NewReader returning an error, gzipReader assigned
a *gzip.Reader-typed nil value to the gz.zr interface value.

On a subsequent Read, gz.zr would not be equal to ==, because it was
actually equal to (type *gzip.Reader, nil), and then zr.Read would call
(*gzip.Reader).Read with a nil receiver and explode.

Debugged internally. (http://go/http2gzipbug)

Change-Id: Icba040ace8ffac3536e5e7ade6695c7660838ca1
Reviewed-on: https://go-review.googlesource.com/19483Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
      cbbbe2bc
  29. 11 Feb, 2016 1 commit
  31. 05 Feb, 2016 2 commits
  33. 03 Feb, 2016 2 commits
  35. 02 Feb, 2016 1 commit
  37. 01 Feb, 2016 2 commits
    • Brad Fitzpatrick's avatar
      http2: don't add *Response to activeRes in Transport on Headers.END_STREA · 644ffc06
      Brad Fitzpatrick authored 
      Prevents a memory leak.

Tests (to be updated) in Go standard library.

Updates golang/go#14084

Change-Id: I3ff602a013bb8fda7a17bccb31beadb08421ae6a
Reviewed-on: https://go-review.googlesource.com/19134Reviewed-by: 's avatarBlake Mizerany <blake.mizerany@gmail.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      644ffc06
    • Brad Fitzpatrick's avatar
      http2: add mechanism to send undeclared Trailers mid handler · d513e585
      Brad Fitzpatrick authored 
      This adds a way for http.Handlers to set trailers after the header has
already been flushed. This comment from the code explains:

// promoteUndeclaredTrailers permits http.Handlers to set trailers
// after the header has already been flushed. Because the Go
// ResponseWriter interface has no way to set Trailers (only the
// Header), and because we didn't want to expand the ResponseWriter
// interface, and because nobody used trailers, and because RFC 2616
// says you SHOULD (but not must) predeclare any trailers in the
// header, the official ResponseWriter rules said trailers in Go must
// be predeclared, and then we reuse the same ResponseWriter.Header()
// map to mean both Headers and Trailers.  When it's time to write the
// Trailers, we pick out the fields of Headers that were declared as
// trailers. That worked for a while, until we found the first major
// user of Trailers in the wild: gRPC (using them only over http2),
// and gRPC libraries permit setting trailers mid-stream without
// predeclarnig them. So: change of plans. We still permit the old
// way, but we also permit this hack: if a Header() key begins with
// "Trailer:", the suffix of that key is a Trailer. Because ':' is an
// invalid token byte anyway, there is no ambiguity. (And it's already
// filtered out) It's mildly hacky, but not terrible.

The official pre-declaring way still works. Example from net/http docs:
https://golang.org/pkg/net/http/#example_ResponseWriter_trailers
And ResponseWriter docs explaining it:
https://golang.org/pkg/net/http/#ResponseWriter

I don't want to add a new interface-assertable upgrade type (like
Hijacker or Flusher) for this because it's hurts composability and
makes everybody in the ecocsystem implement those, and two optional
interfaces is already bad enough. This is a weird enough feature
(Trailers by itself is weird enough), that I don't feel like a third
optional interface is worth it.

This code also filters invalid header fields (updates golang/go#14048),
since I had to update that code as part of this. But I want to later
return an error back to the user if possible. Or log something.

With this CL, all the grpc-go end2end tests pass with a new http2-based
Server implementation:
https://github.com/bradfitz/grpc-go/commit/a06f8f0593bfa8a0af72e57fb3916144f7b30121

Change-Id: I80f863d05a1810bd6f302f34932ad9df0a6646a6
Reviewed-on: https://go-review.googlesource.com/19131Reviewed-by: 's avatarAndrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      d513e585
  39. 26 Jan, 2016 3 commits
  41. 21 Jan, 2016 1 commit
  43. 20 Jan, 2016 2 commits
  45. 19 Jan, 2016 2 commits
  47. 18 Jan, 2016 1 commit
  49. 16 Jan, 2016 1 commit
  51. 15 Jan, 2016 1 commit
    • Brad Fitzpatrick's avatar
      http2: make Transport close unneeded connections after h1->h2 upgrade · a8e212f3
      Brad Fitzpatrick authored 
      If a user starts two HTTP requests when no http2 connection is
available, both end up creating new TCP connections, since the
server's protocol (h1 or h2) isn't yet known. Once it turns out that
the server supports h2, one of the connections is useless. Previously
we kept upgrading both TLS connections to h2 (SETTINGS frame exchange,
etc).  Now the unnecessary connections are closed instead, before the
h2 preface/SETTINGS.

Tests in the standard library (where it's easier to test), in the
commit which updates h2_bundle.go with this change.

Updates golang/go#13957

Change-Id: I5af177e6ea755d572b551cc0b0de9da865ef4ae7
Reviewed-on: https://go-review.googlesource.com/18675Reviewed-by: 's avatarAndrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      a8e212f3