Strange issue while pushing/pulling to Hg repo hosted by SCM-Manager

[Stephen Lorenz]

Hello,

Our development team consists of several developers, several of whom use PC's (Windows 7) and several use Macs (OSX Lion).  Our Windows based developers never seem to experience the issues described below.  We all use the Eclipse IDE and the same MercurialEclipse plugin.  The issue is intermittent but chronic (maybe 80% of the time) and occurs when these Mac developers attempt push or pull from the Mercurial repository managed by SCM-Manager.  When it problem occurs, these users will see an error message similar to the following:

Incoming comand failed for /Volumes/Bravo/Workspace/Workspace-3.7/portal. using http://repo.anydomain.com.:8045/scm/hg/portal
http auth: user jdoe, password *********
sending capabilities command
http auth: user jdoe, password *********
comparing with http://jdoe:***@repo.anydomain.com:8045/scm/hg/portal
query 1; heads
sending batch command
http auth: user jdoe, password *********
searching for changes
taking initial sample
searching: 2 queries
query 2; still undecided: 357, sample size is: 200
sending known command
abort: error: Connection reset by peer. Command line: /Volumes/Bravo/Workspace/Workspace-3.7/portal:hg -y incoming --debug --style /Volumes/Bravo/Workspace/Workspace-3.7/.metadata/.plugins/com.vectrace.MercurialEclipse/log_style_with_files --bundle /private/var/folders/xs/l48xs4996_dbqw4pypyv6l9c0000gn/T/bundleFile-portal-7919336165314085165.tmp http://***@repo.anydomain.com:8045/scm/hg/portal, error code: 255
using http://repo.anydomain.com:8045/scm/hg/portal
http auth: user jdoe, password *********
sending capabilities command
http auth: user jdoe, password *********
comparing with http://jdoe:***@repo.anydomain.com:8045/scm/hg/portal
query 1; heads
sending batch command
http auth: user jdoe, password *********
searching for changes
taking initial sample
searching: 2 queries
query 2; still undecided: 357, sample size is: 200
sending known command
abort: error: Connection reset by peer. Command line: /Volumes/Bravo/Workspace/Workspace-3.7/portal:hg -y incoming --debug --style /Volumes/Bravo/Workspace/Workspace-3.7/.metadata/.plugins/com.vectrace.MercurialEclipse/log_style_with_files --bundle /private/var/folders/xs/l48xs4996_dbqw4pypyv6l9c0000gn/T/bundleFile-portal-7919336165314085165.tmp http://***@repo.anydomain.com:8045/scm/hg/portal

These same Mac users never run into this issue when connecting to the same repo served by the hgweb apache plugin.  (For obvious reasons we would prefer SCM-Manager over hgweb; hgweb is not a long term solution.)  If these Mac users experienced the same problem while connecting to the repo via hgweb, we would assume the problem was perhaps with the Mac version of the MercurialEclipse plugin; but alas, the problem never occurs with hgweb.

We are running SCM-Manager 1.9 on a Windows 2003 Server running Tomcat 6.

Any help or suggestions would be most appreciated!

Thanks!

Stephen 

[Sebastian Sdorra]

Hello,
Which version of mercurial have the mac users? Could you post the
output of the scm-manager log?
The issue could be a problem with a too small header buffer. Could you
please set the maxHttpHeaderSize of your connector to 16384
(http://tomcat.apache.org/tomcat-6.0-doc/config/http.html) and test it
again?

Sebastian

2011/12/28 Stephen Lorenz <stephen...@gmail.com>:

[Stephen Lorenz]

Hi Sebastian,

Thank you for your quick reply.  I will set the maxHttpHeaderSize and have the Mac based developers try again.  I will also capture the  scm-manager log output and send that as well.  Unfortunately, I will have to get back to you next week as our 3 Mac developers are out this week on vacation.

Thanks again,

Stephen