On Thu, Apr 26, 2007 at 10:20:16PM +0800, Jason Zhao wrote:
> How about this:
>
> if [[ $SCM_MODE == "unknown" ]];then
> print -u2 "Unable to determine SCM type currently in use."
> print -u2 "For teamware: please looks for \$CODEMGR_WS either in"
> print -u2 " the environment or in the file list."
> print -u2 "For mercurial: please runs 'hg root'."
> exit 1
"please"? You could put $0 if you want to be generic, or hardcode it to
xref. But I think the logic implied by the text is backwards. I think
it's probably best left to the man page for explanation.
> >>function detect_scm
> >>{
> >> #
> >> # The presence of $CODEMGR_WS and a Codemgr_wsdata directory
> >> # is our clue that this is a teamware workspace.
> >> #
> >> if [[ -n $CODEMGR_WS && -d "$CODEMGR_WS/Codemgr_wsdata" ]]; then
> >>
> >
> >Wouldn't the output of workspace name be more appropriate here?
> >
> I am not sure what you meant?
> output the workspace name?
No, sorry; I meant the output of the command "workspace name".
> I made a test,the time take almost the same for "hg locate" and "hg manifest"
> in the directory of "hg root",see below.
> ######### hg locate "glob:$dir/**/${pat#s.}" #########
> # time flg.flp -r
> ......
> real 3m38.654s
> user 2m41.522s
> sys 0m24.115s
>
> ######### hg manifest | grep "^$dir/.*/${pat#s.}\$" #######
> # time flg.flp -r
> ......
> real 10m11.090s
> user 8m2.345s
> sys 0m45.250s
Hm. I'm not seeing "almost the same" times, but that manifest takes nearly
three times as long. Weird. But then again, I was testing just the hg
commands, and not all of flg.flp, so maybe something else is going on. Do
you get similar differences when just running the hg commands?
> Then I have an idea,why couldn't we establish a snapshot of all filelist when
> the workspace is detected as "mercurial",then try to find the $dir/${pat#s.}
> in
> this snapshot.Is it a better,I am modifying the script to make a test.Like
> this:
Hm. The command is run multiple times. I'd probably just find a way to
not have to do that. Something like
for dir; do
dirs="$dirs|$dir"
done
dirs=${dirs#|}
hg manifest | egrep "^($dirs)/.*/${pat#s.}\$"
or something like that. I dunno if that'd actually do what needs to happen
there.
> >> while read srcfile; do
> >> if [ "$RELPATHS" != y -o $CURTREE = $codemgr_ws ]; then
> >
> >Why the change of adding "-o $CURTREE = $codemgr_ws"?
> >
> This is because when I run "flg.flp -r" at $CODEMGR_WS($CURTREE is
> $CODEMGR_WS),
So this is a pre-existing bug?
> >>if [ "${SUBTREE##$codemgr_ws}" = "$SUBTREE" ]; then
> >> fail "$SUBTREE is not a subtree of Workspace"
> >>fi
> >>
> >>if [ "${CURTREE##$codemgr_ws}" = "$CURTREE" ]; then
> >> fail "$CURTREE is not a subtree of Workspace"
> >>fi
> >>
> >
> >These used to read "$SUBTREE is not a subtree of \$CODEMGR_WS"; presumably
> >this change isn't related to making this work with mercurial, is it? (It
> >may be a perfectly good change, but should probably fall under a different
> >bugid).
> >
> yes,I think it may relate to mercurial.
> If the SUBTREE donot in the "hg root",it will not run without fault.
> Maybe I will change like it.
>
> fail "$SUBTREE is not a subtree of $codemgr_ws"
Right, that's what I'd expect.
Danek
