iojs 1.4.x, linux: ../deps/uv/src/unix/stream.c:1174: uv_shutdown; win: shutdown EPIPE

[qfox]

$ uname -a
Linux dev.qfox.ru 3.2.0-4-486 #1 Debian 3.2.51-1 i686 GNU/Linux

Expected as it was:

$ iojs -v
v1.3.0
$ iojs -e 'process.stdin.emit("end");'

Silent exit.

Actual:

$ iojs -v
v1.4.3
$ iojs -e 'process.stdin.emit("end");'
iojs: ../deps/uv/src/unix/stream.c:1174: uv_shutdown: Assertion `(stream->type == UV_TCP || stream->type == UV_NAMED_PIPE) && "uv_shutdown (unix) only supports uv_handle_t right now"' failed.
Аварийный останов

cc @indutny

upd Btw, this code also produces strange error on win7 x64 (running in mingw32):

$ iojs -v
v1.4.3
$ iojs -e 'process.stdin.emit("end");'
events.js:141
      throw er; // Unhandled 'error' event
            ^
Error: shutdown EPIPE
    at exports._errnoException (util.js:734:11)
    at ReadStream.onSocketFinish (net.js:218:26)
    at emitNone (events.js:67:13)
    at ReadStream.emit (events.js:163:7)
    at finishMaybe (_stream_writable.js:477:14)
    at endWritable (_stream_writable.js:486:3)
    at ReadStream.Writable.end (_stream_writable.js:452:5)
    at ReadStream.Socket.end (net.js:393:31)
    at process._tickCallback (node.js:350:11)

[bnoordhuis]

Confirmed. Do you happen to know if older io.js versions are affected? If not, I'll just run git bisect.

[qfox]

@bnoordhuis Affected versions are 1.4.1, 1.4.2, 1.4.3.
1.3.0 is fine.

[bnoordhuis]

@zxqfox Thanks. I'll have a look.

[Fishrock123]

Also happens on OS X. (Same error as Linux)

[bnoordhuis]

b9686233fc0be679d7ba1262b611711629ee334e is the first bad commit
commit b9686233fc0be679d7ba1262b611711629ee334e
Author: Fedor Indutny <[email protected]>
Date:   Sun Feb 22 21:59:07 2015 +0300

    stream_base: introduce StreamBase

    StreamBase is an improved way to write C++ streams. The class itself is
    for separting `StreamWrap` (with the methods like `.writeAsciiString`,
    `.writeBuffer`, `.writev`, etc) from the `HandleWrap` class, making
    possible to write abstract C++ streams that are not bound to any uv
    socket.

    The following methods are important part of the abstraction (which
    mimics libuv's stream API):

    * Events:
      * `OnAlloc(size_t size, uv_buf_t*)`
      * `OnRead(ssize_t nread, const uv_buf_t*, uv_handle_type pending)`
      * `OnAfterWrite(WriteWrap*)`
    * Wrappers:
      * `DoShutdown(ShutdownWrap*)`
      * `DoTryWrite(uv_buf_t** bufs, size_t* count)`
      * `DoWrite(WriteWrap*, uv_buf_t*, size_t count, uv_stream_t* handle)`
      * `Error()`
      * `ClearError()`

    The implementation should provide all of these methods, thus providing
    the access to the underlying resource (be it uv handle, TLS socket, or
    anything else).

    A C++ stream may consume the input of another stream by replacing the
    event callbacks and proxying the writes. This kind of API is actually
    used now for the TLSWrap implementation, making it possible to wrap TLS
    stream into another TLS stream. Thus legacy API calls are no longer
    required in `_tls_wrap.js`.

    PR-URL: https://github.com/iojs/io.js/pull/840
    Reviewed-By: Trevor Norris <[email protected]>
    Reviewed-By: Chris Dickinson <[email protected]>

/cc @indutny

[indutny]

Ok, on my plate then :)

[indutny]

Should be fixed by #1073.

[indutny]

Fixed!

[Fishrock123]

@indutny the test case for this is still failing: win2012r2 & win2008r2

[indutny]

Gosh, cc @piscisaureus: PTAL

[qfox]

Just fyi: I can't repeat it on my local win machine with v1.4.4-nightly20150305b27931b0fe

[indutny]

Neither do I.

[rvagg]

@indutny, @piscisaureus: the failure of test-regress-GH-io-1068 is the last big item in #1005 that we need to sort out for Windows. The behaviour exhibits when run from within Jenkins but does not when you run the tests or the command directly with iojs -e ... I'm guessing it's the fact that when run from under Jenkins it's not actually a TTY so kFlagNoShutdown is not set on that stream, but I have no idea what else it would be if not a TTY and why this would be a problem on Windows. Is there another type of system stream that's not happy with the shutdown?

[indutny]

@rvagg is there any way to check if the test was failing before the streambase stuff? If so - we may want to just disable it.

[Fishrock123]

@indutny looks like @bnoordhuis pointed out above in #1068 (comment) that it originated from stream_base..

[indutny]

@Fishrock123 yeah, but it the unintentional behavior change was fixed since then. So practically io.js should behave the same way as it was doing before the StreamBase. That's why I'm asking about reproducing it on previous io.js versions.

[rvagg]

@indutny you could make a branch in your own repo from just before StreamBase and put in that regress test and run it through CI and that'll tell you ... you could do an awkward bisect using that method too I suppose, if you have patience.

[indutny]

Here it is, right before the StreamBase: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/348/ (crossing my fingers :) )

[rvagg]

IMO even if this does turn out to exist prior to StreamBase it's now something that we have to find a way to deal with, it's just a bug that was previously unexposed. I'd like some thought put in to what conditions are causing this and how to "fix" it so that it doesn't throw.

[indutny]

@rvagg sure thing. I'll probably need to spend some time reproducing it, though.

[rvagg]

I spent some time yesterday trying to repro it with child_process gymnastics to no avail, perhaps we can use python to invoke something similar to what's going on with Jenkins

[indutny]

And the build is blue, right before the StreamBase. I'll see what I could have broke there.

[indutny]

Yay, I was able to reproduce it:

echo 1 | ./Release/iojs.exe test/parallel/test-regress-GH-io-1068.js

Will figure out the problem in a bit.

[indutny]

Good news, test appears to be failing right before the StreamBase changes too. Perhaps this is still unrelated to the stuff that we see on Jenkins.

[indutny]

@rvagg: windows CI failures should be resolved now.