Re: Minor VCS Info things

[Peter Stephenson <pws@xxxxxxx>]

On Mon, 22 Sep 2008 15:11:02 +0100
Peter Stephenson <pws@xxxxxxx> wrote:
> The Perforce detector doesn't work very well.  $P4CONFIG is not tied to
> being in the Perforce workspace: you can have a config file outside a
> workspace to set a default client etc., and you don't need one at all if
> the client etc. are set via environment variables.  The only good way of
> detecting whether you're in a workspace is to ask the server.
> Unfortunately this can hang (though not permanently) if the server is
> not available, so I haven't sent a patch.  It could be fixed by adding
> an explicit style to use the server and also a global (per server:port
> pair) flag that the server didn't respond, so that it only hangs (for a
> second or two) the first time.

I've done this.  It seems to work OK.

Index: Doc/Zsh/contrib.yo
===================================================================
RCS file: /cvsroot/zsh/zsh/Doc/Zsh/contrib.yo,v
retrieving revision 1.88
diff -u -r1.88 contrib.yo
--- Doc/Zsh/contrib.yo	22 Sep 2008 17:22:48 -0000	1.88
+++ Doc/Zsh/contrib.yo	23 Sep 2008 15:55:51 -0000
@@ -502,6 +502,21 @@
 repositories (checked in the var(-init-) context, too). Only used if
 tt(enable) contains tt(ALL).
 )
+kindex(use-server)
+item(tt(use-server))(
+This is used by the Perforce backend (tt(p4)) to decide if it should
+contact the Perforce server to find out if a directory is managed
+by Perforce.  This is the only reliable way of doing this, but runs
+the risk of a delay if the server name cannot be found.  If the
+server (more specifically, the var(host)tt(:)var(port) pair describing the
+server) cannot be contacted its name is put into the associative array
+tt(vcs_info_p4_dead_servers) and not contacted again during the session
+until it is removed by hand.  If you do not set this style, the tt(p4)
+backend is only usable if you have set the environment variable
+tt(P4CONFIG) to a file name and have corresponding files in the root
+directories of each Perforce client.  See comments in the function
+tt(VCS_INFO_detect_p4) for more detail.
+)
 kindex(use-simple)
 item(tt(use-simple))(
 If there are two different ways of gathering
Index: Functions/VCS_Info/Backends/VCS_INFO_detect_p4
===================================================================
RCS file: /cvsroot/zsh/zsh/Functions/VCS_Info/Backends/VCS_INFO_detect_p4,v
retrieving revision 1.1
diff -u -r1.1 VCS_INFO_detect_p4
--- Functions/VCS_Info/Backends/VCS_INFO_detect_p4	19 Sep 2008 12:58:53 -0000	1.1
+++ Functions/VCS_Info/Backends/VCS_INFO_detect_p4	23 Sep 2008 15:55:51 -0000
@@ -2,8 +2,76 @@
 ## perforce support by: Phil Pennock
 ## Distributed under the same BSD-ish license as zsh itself.
 
-[[ -n ${P4CONFIG} ]] || return 1
-VCS_INFO_check_com p4 || return 1
-vcs_comm[detect_need_file]="${P4CONFIG}"
-VCS_INFO_bydir_detect .
-return $?
+# If user-server is true in the :vcs_info:p4:... context, contact the
+# server to decide whether the directory is handled by Perforce.  This can
+# cause a delay if the network times out, in particular if looking up the
+# server name failed.  Hence this is not the default.  If a timeout
+# occurred, the server:port pair is added to the associative array
+# vcs_info_p4_dead_servers and the server is never contacted again.  The
+# array must be edited by hand to remove it.
+#
+# If user-server is false or not set, the function looks to see if there is
+# a file $P4CONFIG somewhere above in the hierarchy.  This is far from
+# foolproof; in fact it relies on you using the particular working practice
+# of having such files in all client root directories and nowhere above.
+
+
+VCS_INFO_p4_get_server() {
+  emulate -L zsh
+  setopt extendedglob
+
+  local -a settings
+  settings=(${(f)"$(p4 set)"})
+  serverport=${${settings[(r)P4PORT=*]##P4PORT=}%% *}
+  case $serverport in
+    (''|:)
+    serverport=perforce:1666
+    ;;
+
+    (:*)
+    serverport=perforce${serverport}
+    ;;
+
+    (*:)
+    serverport=${serverport}1666
+    ;;
+
+    (<->)
+    serverport=perforce:${serverport}
+    ;;
+  esac
+}
+
+
+VCS_INFO_detect_p4() {
+  local serverport p4where
+
+  if zstyle -t ":vcs_info:p4:${usercontext}:${rrn}" use-server; then
+    # Use "p4 where" to decide whether the path is under the
+    # client workspace.
+    if (( ${#vcs_info_p4_dead_servers} )); then
+      # See if the server is in the list of defunct servers
+      VCS_INFO_p4_get_server
+      [[ -n $vcs_info_p4_dead_servers[$serverport] ]] && return 1
+    fi
+    if p4where="$(p4 where 2>&1)"; then
+      return 0
+    fi
+    if [[ $p4where = *"Connect to server failed"* ]]; then
+      # If the connection failed, mark the server as defunct.
+      # Otherwise it worked but we weren't within a client.
+      typeset -gA vcs_info_p4_dead_servers
+      [[ -z $serverport ]] && VCS_INFO_p4_get_server
+      vcs_info_p4_dead_servers[$serverport]=1
+    fi
+    return 1
+  else
+    [[ -n ${P4CONFIG} ]] || return 1
+    VCS_INFO_check_com p4 || return 1
+    vcs_comm[detect_need_file]="${P4CONFIG}"
+    VCS_INFO_bydir_detect .
+    return $?
+  fi
+}
+
+VCS_INFO_detect_p4 "$@"

-- 
Peter Stephenson <pws@xxxxxxx>                  Software Engineer
CSR PLC, Churchill House, Cambridge Business Park, Cowley Road
Cambridge, CB4 0WZ, UK                          Tel: +44 (0)1223 692070