crystal-lang/crystal

The Crystal programming language | http://crystal-lang.org | Fund Crystal's development: http://is.gd/X7PRtI | Docs: http://crystal-lang.org/docs/ | API: http://crystal-lang.org/api/

Isaac Sloan

@elorest

<oprypin> For some reason if the body is even moderately large(~200 elements) it changes to chunked mode and overwrites the existing headers with

< HTTP/1.1 200 OK
< Connection: keep-alive
< Transfer-Encoding: chunked

If I reduce the size of the http body all of my headers get sent.

And my cookies get set in my browser.

From IRC (bridge bot)

@FromIRC

<shelvacu> Is it a consistent size that changes it? Maybe 4k bytes or something?

Isaac Sloan

@elorest

Working one with smaller body.

< HTTP/1.1 200 OK
< Connection: keep-alive
< Set-Cookie: test_session=eyJjc3JmLnRva2VuIjoiSlFMdGVzTXFmUzBwZDlXYmVSSkpMVUtmbGFzME93eEFINkNrRHlXYkNUMCIsIl9mbGFzaCI6IntcImZsYXNoZXNcIjp7fSxcImRpc2NhcmRcIjpbXX0ifQ%3D%3D--OPgRdhc%2BbD0AQp1Xz88kvrQ%2Fwj4%3D; path=/; HttpOnly
< Content-Length: 6680

Same url.

I’ll try to find the content lenght where it breaks.

From IRC (bridge bot)

@FromIRC

<shelvacu> Do you have some code that reliably reproduces it?

Isaac Sloan

@elorest

Yes.

From IRC (bridge bot)

@FromIRC

<shelvacu> You should definitely file a bug report.

<shelvacu> Or post the code here so I can file such a report

Isaac Sloan

@elorest

Content-Length: 8191 works but one more character in the body and it goes to chunked and doesn’t send headers.

From IRC (bridge bot)

@FromIRC

<shelvacu> https://github.com/crystal-lang/crystal/blob/f026c474450176ff1b86293bf0abc6afdc497137/src/http/common.cr#L157 seems the most likely culprit but I don't see how that would cause any problems.

<shelvacu> deflate compression lib and stdlib buffered io both use a buffer size of 8192

<shelvacu> 1

Isaac Sloan

@elorest

Yeah I’m finding stuff about that number everywhere. As far as html pages are concerned it’s not really that big.

I believe that chunking after that size is appropriate. I just tested a couple rails sites. They also chuck after 8191 but they don’t delete my headers in the process.

From IRC (bridge bot)

@FromIRC

<bmcginty> @oprypin: It's with sockets that I'm having the issue. Straight reading from them times out (while checking to see if .read returns 0).

Bar Hofesh

@bararchy

bmcginty: Can you share a code exmaple ?

From IRC (bridge bot)

@FromIRC

<oprypin> bmcginty, no difference

<oprypin> sockets when explicitly closed will get you 0. but detecting remote socket going away is not reliable, hence timeout

From IRC (bridge bot)

@FromIRC

<bmcginty> @oprypin: thanks.

Andrés Pérez

@code-andres

Hi, everyone... I'm stuck using concurrency for a simple logger...

I have a Process, a spawn where the output is print and i want to use another spawn where each 3 seconds for example, kill the old Process and send the new one to the first spawn, I was trying with channels but the spawn must wait for a send.

In little words, is there a way to create just a "listener"between spawns without wait for it?

Bar Hofesh

@bararchy

Buffered Channel maybe ?

Cris Ward

@crisward

Can anyone explain to me what this does

Signal::PIPE.trap do
  # some code
end

Akzhan Abdulin

@akzhan

This is signal handler. They interrupts process execution for some events.

Many operating systems allow signals to be sent to running processes. Some signals have a defined effect on the process, while others may be trapped at the code level and acted upon. For example, your process may trap the USR1 signal and use it to toggle debugging, and may use TERM to initiate a controlled shutdown.

this is citation from Ruby doc :)

And usually should to set ignore on PIPE signal.

Cris Ward

@crisward

@akzhan Thanks... I was using this around a socket.send as I was getting application crashes when the remote end disconnected mid message. However it seems to stop the sockets sending altogether. I can't remember what the exception was.

I think someone on here suggested it, I added it a while ago, but just realised it's actually preventing the sockets being sent.

From IRC (bridge bot)

@FromIRC

<oprypin> @crisward, putting code inside signal.trap doesn't mean wrapping it, the code actually won't run. the code will run when the corresponding signal is received

<oprypin> so the suggestion to use that probably meant something different

Cris Ward

@crisward

oprypin - so it's like a rescue on begin, rescue ? Should I therefore just do this once globally ? Or once before every socket.send ? Perhaps I'll take a look at the ruby docs.

From IRC (bridge bot)

@FromIRC

<oprypin> once globally

<oprypin> @crisward, https://crystal-lang.org/api/master/Signal.html

<oprypin> https://crystal-lang.org/api/master/Signal.html#trap%28%26block%3ASignal-%3E%29-instance-method

<oprypin> not yet released docs

Cris Ward

@crisward

@oprypin Thanks! Never thought of looking in the docs on master.

From IRC (bridge bot)

@FromIRC

<oprypin> well sure, there's almost never a reason for that

From IRC (bridge bot)

@FromIRC

<RX14> i got a segfault in the compiler while investigating a segfault in a program

<RX14> which is nice

<RX14> https://gist.github.com/RX14/b468ea7cd78dddcc004ffa704b5238cf

From IRC (bridge bot)

@FromIRC

<RX14> my segfault disappears under gdb :⁠ok_hand:

From IRC (bridge bot)

@FromIRC

<RX14> so somehow this generated code is ignoring the calling conventions and overwriting rbx

<RX14> ok so it's stack corruption

<RX14> damnit

From IRC (bridge bot)

@FromIRC

<oprypin> sounds pleasant

From IRC (bridge bot)

@FromIRC

<RX14> @oprypin, read man 3 accept and tell me i'm crazy to think that using out here is completely wrong: https://github.com/crystal-lang/crystal/blob/master/src/socket.cr#L219

<RX14> man page says you should allocate struct sockaddr*

Where communities thrive

People

Repo info

Activity