Bug #6436: Byte counting incorrect in BufferedIO (net/protocol.rb)

• Status: Closed
• Assignee: nobu (Nobuyoshi Nakada)
• Target version: -
• ruby -v: 1.9

[jcast (Jeremie Castagna)]

BufferedIO's @rbuf counts bytes with String#size, but should use String#bytesize. This creates potential incorrect reading of a HTTP body when changing the encoding of each chuck passed by HTTP#read_body:

http.read_body do |chunk|
chunk.force_encoding "UTF-8"
end

[jcast (Jeremie Castagna)]

• File protocol.patch protocol.patch added

Patch file.

[nobu (Nobuyoshi Nakada)]

• File protocol.patch protocol.patch added

Emacs diff-mode.el provides a function, diff-context->unified.

[mame (Yusuke Endoh)]

• Status changed from Open to Assigned
• Assignee set to nobu (Nobuyoshi Nakada)

Thanks Jeremie,
Could you add a test to test/net/http/test_buffered_io.rb ?

Nobu, can you review the patch?

--
Yusuke Endoh [e-mail]

[jeremyevans0 (Jeremy Evans)]

• Status changed from Assigned to Closed

It looks like this issue was fixed between Ruby 2.4 and 2.5:

$ ruby24 -rnet/http -e "p Net::HTTP.start('www.falcom.co.jp', 443, :use_ssl=>true){|http| http.request_get('/'){|res| res.read_body{|c| p c.bytesize;  c.force_encoding 'UTF-8'}}}" 
0
16384
16384
16384
4604
2
5
0
16384
16384
16384
4604
2
5
-e: end of file reached (EOFError)

$ ruby24 -rnet/http -e "p Net::HTTP.start('www.falcom.co.jp', 443, :use_ssl=>true){|http| http.request_get('/'){|res| res.read_body{|c| p c.bytesize}}}"                            
0
16384
16384
16384
4604
#<Net::HTTPOK 200 OK readbody=true>

$ ruby25 -rnet/http -e "p Net::HTTP.start('www.falcom.co.jp', 443, :use_ssl=>true){|http| http.request_get('/'){|res| res.read_body{|c| p c.bytesize;  c.force_encoding 'UTF-8'}}}" 
0
16384
16384
16384
4604
#<Net::HTTPOK 200 OK readbody=true>