[cvsnt] CVSNT client hangs when using compression with 1.11.2 UNIX pserve r

[Torsten Martinsen]

Other users have reported the same behaviour with CVSNT build 57h and 57i.

Bo, are your tests also against a _UNIX_ server (not CVSNT on Unix, but "plain" CVS)?

-Torsten

----- Original Message ----- 
From: "Bo Berglund" <bo.berglund at telia.com>
Newsgroups: support.cvsnt
To: <cvsnt at cvsnt.org>
Sent: Tuesday, October 08, 2002 12:53 AM
Subject: Re: [cvsnt] CVSNT client hangs when using compression with 1.11.2 UNIX pserve r


> Do yourself a favour and check out the branch:
> CVSNT_1_11_1_3_REL
> instead of HEAD.
> There are all kinds of strange stuff going on at HEAD, it is really
> alpha code (as witnessed by the version output). Don't expect it to
> work problem free.
> The REL branch is much safer because Tony only sends updates there
> after rather a lot of testing.
> 
> /Bo
> 
> BTW: My tests with build 57h work just fine.
> 
> 
> On Mon, 7 Oct 2002 16:25:23 -0400, Elliot Murphy
> <elliot.murphy at veritas.com> wrote:
> 
> >Hello,
> > 
> >I am seeing a problem where the CVSNT client blocks at buf_shutdown when
> >connected to a UNIX pserver. I initially started seeing this problem when
> >one of the CVS servers that I connect to upgraded to CVS 1.11.2. I have seen
> >this with a number of different versions of the CVSNT client, and have
> >verified that it does not happen with the 1.11.2 Win32 client that is
> >available from www.cvshome.org <http://www.cvshome.org> . In order to get
> >more information about what is happening, I checked out the latest cvsnt
> >source a couple of hours ago and built a debug version of the client.
> > 
> >I am running the command:
> >cvs -z 3 version
> > 
> >The output is this:
> >Client: Concurrent Versions System (CVSNT) 1.11.1.4 Alpha 2 (Nightly Build
> >Debug Build 61) (client/server)
> >Server: Concurrent Versions System (CVS) 1.11.2 (client/server)
> > 
> >At this point the client is in get_responses_and_close(). The call to
> >buf_shutdown(to_server) appears to go through just fine, and then the call
> >to buf_shutdown(from_server) blocks. I have traced into
> >buf_shutdown(from_server), and it is blocking in a recv() call. The call
> >stack looks like this:
> > 
> >   <mailto:wsock32.dll!_recv at 16> wsock32.dll!_recv at 16()  
> >  pserver_protocol.dll!tcp_read_raw(void * data=0x00af0000, int
> >length=0x00001000)  Line 205 + 0x19 C
> >  pserver_protocol.dll!pserver_read_data(const protocol_interface *
> >protocol=0x1001a550, void * data=0x00af0000, int length=0x00001000)  Line
> >278 + 0x10 C
> >  cvs.exe!client_buffer_input(void * closure=0x003a78a0, char *
> >data=0x00af0000, int need=0x00000001, int size=0x00001000, int *
> >got=0x0012f3f0)  Line 567 + 0x18 C
> >  cvs.exe!buf_read_line(buffer * buf=0x003a78e0, char * * line=0x0012f554,
> >int * lenp=0x0012f548)  Line 818 + 0x1d C
> >  cvs.exe!read_line(char * * resultp=0x0012f65c)  Line 640 + 0x14 C
> >  cvs.exe!get_server_responses()  Line 3240 + 0x9 C
> >  cvs.exe!start_server(int verify_only=0x00000000)  Line 3627 + 0x5 C
> >  cvs.exe!cvsguiglue_main(int argc=0x00000001, char * * argv=0x003a635c)
> >Line 1126 + 0x7 C
> >  cvs.exe!main(int argc=0x00000003, char * * argv=0x003a6040)  Line 222 +
> >0x2a C
> >
> >The problem only happens when I use compression, which is consistent with
> >the comments near buf_shutdown(). I have no control over the CVS server that
> >I am connecting to, and I doubt the sysadmin there will be very sympathetic
> >to me since the "plain" 1.11.2 cvs client does not exhibit this behavior.
> > 
> >Any ideas about what might be causing this problem? I'm happy to debug
> >further if more information would be helpful.