Re: [PATCH 4/4] mergetools/meld: Use '--output' when available
- From
David Aguilar <davvid@gmail.com>
- Date
- Aug 18, 2011, 09:24 UTC
- Message-ID
- <20110818092403.GB15416@gmail.com>
- In-Reply-To
- <20110818081309.GP31888@elie.gateway.2wire.net>
On Thu, Aug 18, 2011 at 03:13:10AM -0500, Jonathan Nieder wrote:
Show 19 quoted lines
> David Aguilar wrote:
>
> > use the '--output' option when available.
>
> Yay. :)
>
> > --- a/mergetools/meld
> > +++ b/mergetools/meld
> > @@ -4,6 +4,37 @@ diff_cmd () {
> >
> > merge_cmd () {
> > touch "$BACKUP"
> > - "$merge_tool_path" "$LOCAL" "$MERGED" "$REMOTE"
> > + if test "$meld_has_output_option" = true
> > + then
> > + "$merge_tool_path" --output "$MERGED" \
> > + "$BASE" "$LOCAL" "$REMOTE"
>
> Shouldn't this be "$LOCAL" "$BASE" "$REMOTE"?Yup, thanks.
Show 19 quoted lines
> > > + else > > + "$merge_tool_path" "$LOCAL" "$MERGED" "$REMOTE" > > + fi > > check_unchanged > > I wonder if the version test could be made a little simpler (perhaps > to cope better if future versions use a different numbering system): > > if "$merge_tool_path" --output /dev/null --help >/dev/null 2>&1 > then > "$merge_tool_path" --output ... > else > "$merge_tool_path" "$LOCAL" ... > fi > > Forgive my ignorance: is this function likely to be called in a loop? > If so, it makes sense to precompute or cache the result of detection, > like you already do.
Yes, it is called in a loop...
Show 50 quoted lines
> check_meld_for_output_option () {
> if ...
> then
> meld_has_output_option=true
> else
> meld_has_output_option=false
> fi
> }
>
> merge_cmd () {
> if test -z "${meld_has_output_option:+set}"
> then
> check_meld_for_output_option
> fi
>
> if test "$meld_has_output_option" = true
> then
> ...
>
>
> [...]
> > + # Filter meld --version to contain numbers and dots only
> > + meld="$(meld --version 2>/dev/null | sed -e 's/[^0-9.]\+//g')"
>
> \+ is not a BRE. If parsing version numbers seems like the right
> thing to do, maybe "tr -cd 0-9."?
>
> > + meld="${meld:-0.0.0}"
> > +
> > + meld_major="$(expr "$meld" : '\([0-9]\{1,\}\)' || echo 0)"
> > + meld_minor="$(expr "$meld" : '[0-9]\{1,\}\.\([0-9]\{1,\}\)' || echo 0)"
>
> I think git avoids \{m,n\} ranges where possible (for portability).
> This could be:
>
> meld_major=${meld%%.*}
> meld_nonmajor=${meld#${meld_major}.}
> meld_minor=${meld_nonmajor%%.*}
>
> or:
>
> case $meld in
> [2-9].* | [1-9][0-9]* | 1.[5-9]* | 1.[1-9][0-9]*) # >= 1.5.0
> meld_has_output_option=true ;;
> *)
> meld_has_output_option=false ;;
> esac
>
> It's nice how self-contained this can be now that it's in its own
> file. Thanks.Right, I was using \{1,\} since that's what CodingStyle said to use instead of \+ (which I forgot to fixup above as you saw).
The case statement is nice and simple enough to understand. By doing meld --output /dev/null --help we're relying on older versions blowing up with --output and --help returning exit status 0. That seems pretty reasonable. The case statement does seem sufficient but just trying --output and seeing what happens is even simpler. Your example also uses $merge_tool_path, which I forgot to do, so I'll be sure to include that too.
I'll wait until tomorrow to see if there are any more comments and reroll.
Thanks,
-- David