Skip Menu |

This queue is for tickets about the Net-HTTP CPAN distribution.

Report information
The Basics
Id: 115721
Status: resolved
Priority: 0/
Queue: Net-HTTP
People
Owner: Nobody in particular
Requestors: martin.lakov [...] sap.com
Cc:
AdminCc:
Bug Information
Severity: (no value)
Broken in: (no value)
Fixed in: (no value)

History Show all quoted text
  Wed Jun 29 04:55:06 2016 martin.lakov [...] sap.com - Ticket created
CC: "Schmoger, Daniel" <daniel.schmoger [...] sap.com>, "Andreev, Ivan" <ivan.andreev [...] sap.com>, "Voelkel, Robert" <robert.voelkel [...] sap.com>
Subject: Bug in Net::HTTP::Methods::my_readline , version 6.09
Date: Wed, 29 Jun 2016 08:54:49 +0000
To: "bug-Net-HTTP [...] rt.cpan.org" <bug-Net-HTTP [...] rt.cpan.org>
From: "Lakov, Martin" <martin.lakov [...] sap.com>
Hello, I would like to inform you about a problem related to the implementation of the following sub Net::HTTP::Methods::my_readline in version 6.09 of the module. Currently the implementation of this sub is "greedy" - it tries to read all available bytes in the byte stream for the request, instead of stopping after a new line was found (which the implementation in 6.06 and 6.07 did). This behavior results in the fact that when we have a continuous feed of data (e.g. downloading large binary file), all of the data is stored in memory and no handlers are notified until the entire payload of the request is received. This can result in an unpredictable behavior (segmentation faults occurred sporadically) and "out-of-memory" errors, depending on the underlying hardware. Best Regards, Martin
  Wed Jun 29 14:43:00 2016 ether [...] cpan.org - Correspondence added
On 2016-06-29 01:55:06, martin.lakov@sap.com wrote: Show quoted text
> Hello, > > I would like to inform you about a problem related to the > implementation of the following sub Net::HTTP::Methods::my_readline in > version 6.09 of the module. > Currently the implementation of this sub is "greedy" - it tries to > read all available bytes in the byte stream for the request, instead > of stopping after a new line was found (which the implementation in > 6.06 and 6.07 did). > This behavior results in the fact that when we have a continuous feed > of data (e.g. downloading large binary file), all of the data is > stored in memory and no handlers are notified until the entire payload > of the request is received. > This can result in an unpredictable behavior (segmentation faults > occurred sporadically) and "out-of-memory" errors, depending on the > underlying hardware. > > Best Regards, > Martin
This looks to be a side effect of the fix made in RT#104122.
  Wed Jun 29 14:43:00 2016 The RT System itself - Status changed from 'new' to 'open'
  Wed Jun 29 14:43:14 2016 ether [...] cpan.org - Reference to ticket #104122 added
  Mon Dec 19 18:18:10 2016 bash [...] cpan.org - Correspondence added
Have the same problem. LWP::UserAgent->get( ... , :content_cb => \&callback) doesnt call callback() because of this bug in my_readline().
  Wed Dec 21 10:13:28 2016 mods [...] hank.org - Correspondence added
Must be related to https://rt.cpan.org/Ticket/Display.html?id=112313 as well.
  Wed Dec 21 10:13:34 2016 mods [...] hank.org - Correspondence added
Must be related to https://rt.cpan.org/Ticket/Display.html?id=112313 as well.
  Sat Feb 11 10:55:07 2017 bash [...] cpan.org - Correspondence added
Fixed for me from 6.10
  Sat Feb 11 14:36:53 2017 olaf [...] wundersolutions.com - Correspondence added
On Sat Feb 11 10:55:07 2017, BASH wrote: Show quoted text
> Fixed for me from 6.10
Great! Thanks for letting us know.
  Sat Feb 11 14:37:04 2017 olaf [...] wundersolutions.com - Status changed from 'open' to 'resolved'