Re: [Libevent-users] Zlib inclusion (and regress failures)

2011-12-19 Thread grarpamp
> Right. It's only used in the regression tests.  There should probably > be a way to disable it entirely, if you'd rather avoid the whole > thing.  As it is, there's no point in pointing it at a different > version. Thanks. Figured it wasn't likely part of the resultant libs. I did look through t

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Mark Ellzey
On Mon, Dec 19, 2011 at 09:57:08PM -0500, Nick Mathewson wrote: > I agree with the part of your fix where we don't ignore the extra data > whose presence is indicated by a nonzero SSL_pending. > > But I think these fixes all have some problem(s) in common: They > ignore the reason for the logic

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Nick Mathewson
On Mon, Dec 19, 2011 at 8:41 PM, Mark Ellzey wrote: > On Mon, Dec 19, 2011 at 07:34:50PM -0600, Mark Ellzey wrote: >> I think I have been able to reproduce this using ssl client + filters. >> More on this in a bit. > > > This seems to fix the issue: > > -               n_to_read = SSL_pending(bev_

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Mark Ellzey
On Mon, Dec 19, 2011 at 07:34:50PM -0600, Mark Ellzey wrote: > I think I have been able to reproduce this using ssl client + filters. > More on this in a bit. I reproduced by running: sample/le-proxy -s 127.0.0.1:8080 127.0.0.1:443 where the 443 host is a https server. Made a 2GB POST, without the

Re: [Libevent-users] Zlib inclusion (and regress failures)

2011-12-19 Thread Nick Mathewson
On Mon, Dec 19, 2011 at 8:33 PM, grarpamp wrote: > Hi. I notice this output from configure... > > checking zlib.h usability... yes > checking zlib.h presence... yes > checking for zlib.h... yes > checking for library containing inflateEnd... -lz > > As such, where is zlib used in libevent? The onl

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Mark Ellzey
On Mon, Dec 19, 2011 at 07:34:50PM -0600, Mark Ellzey wrote: > I think I have been able to reproduce this using ssl client + filters. > More on this in a bit. This seems to fix the issue: - n_to_read = SSL_pending(bev_ssl->ssl); + if (!(n_to_read = SSL_pending(bev_ssl

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Mark Ellzey
On Mon, Dec 19, 2011 at 05:45:51PM -0500, Catalin Patulea wrote: > > Ps - I haven't tried to reproduce this bug in a stand alone project yet. > Would you be able to provide a repro, or at least some more in-depth > description of the conditions that prompted the issue to surface? I think I have b

[Libevent-users] Zlib inclusion (and regress failures)

2011-12-19 Thread grarpamp
Hi. I notice this output from configure... checking zlib.h usability... yes checking zlib.h presence... yes checking for zlib.h... yes checking for library containing inflateEnd... -lz As such, where is zlib used in libevent? The only place I can see is in the regress tests (not a part of the act

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Catalin Patulea
On Mon, Dec 19, 2011 at 5:34 PM, Haseeb Abdul Qadir wrote: > Thanks for resending this on the mailing list. > SSL_pending() ends up returning 0 in some cases even when the underlying BE's > input buffer has data left to consume This is exactly the condition I was afraid of back when 2aa036f was

[Libevent-users] Re: PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Haseeb Abdul Qadir
Also, corresponding pull request: https://github.com/libevent/libevent/pull/2 Thanks. On Tuesday, 20 December, 2011 at 3:34 AM, Haseeb Abdul Qadir wrote: > > > Hi, > > I was investigating an issue with filtered openssl buffer events. The > bufferevent stopped processing / pumping SSL data

[Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-19 Thread Haseeb Abdul Qadir
Hi, I was investigating an issue with filtered openssl buffer events. The bufferevent stopped processing / pumping SSL data event when there was data available in the underlying BE. This is a regression in 2.0.16 and I believe the follow commit might have caused it:https://github.com/libeven