[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [libmicrohttpd] Bug report - chunked transfer encoding not working w

From: Christian Grothoff
Subject: Re: [libmicrohttpd] Bug report - chunked transfer encoding not working with SSL (libgnutls).
Date: Tue, 16 Jun 2020 20:49:26 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.8.0

Dear Philip,

Thanks for your report.

I've done some manual inspection of the code, and -- without having
reproduced the issue -- my best guess is that our ignoring of the result
from gnutls_record_uncork() likely is "unsavory". I have addressed this
in Git commit 11ed3de3..0d771f77 and would very much appreciate you
trying it out.

If that does help, please report it.

If it does NOT help, some minimum example to reproduce the issue would
definitively help me investigate -- after all there could also be ways
how you might be using the API incorrectly, and I'd prefer not to go on
too wild a goose chase ;-).

Happy hacking!


On 6/15/20 10:39 AM, Philip Couling wrote:
> Hi
> I'm seeing a weird bug that's starting to look a lot like it's in
> libmicrohttpd rather than my own code.
> When sending a response via chunked transfer
> (MHD_create_response_from_callback size=-1) the response to the client
> does not seem to complete correctly.  It only does this with HTTPS, when
> I switch to HTTP the bug goes away.  
> I've tested my app with wget and consistently get the error (where the
> response is 122488 long): Read error at byte 122488 (Success).Giving up
> This suggests to me the chunked transfer encoding may be somehow missing
> the terminating 0\r\n
> I'm seeing it in code I wrote a few years ago when run against Archlinux
> 's package version 0.9.7-3.  The program has behaved correctly for years
> and still works under Ubuntu 20.04 with libmicrohttpd version 0.9.66.  I
> have of course checked very carefully that my own code is behaving
> "normally".  IE my program's callback is responding identically with and
> without SSL and the same as it has in previous versions.
> For reference the original bug was reported against my code
> here:  I'm building
> an SSCCE to demonstrate the bug but it's taking me a little longer than
> I'd like, hence I'm sending this bug report without it.
> Kind Regards

Attachment: signature.asc
Description: OpenPGP digital signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]