Re: What is the difference between > and >> in a makefile.

[Stephan Beal]

On Mon, Mar 28, 2011 at 9:55 PM, givemecode <address@hidden> wrote:

>
>
>
> Paul Smith-20 wrote:
> >
> > On Mon, 2011-03-28 at 11:15 -0700, givemecode wrote:
> >> I modified a for loop in a makefile that looked like this:
> >>
> >>      @for f in $(DRV_GEN_HEADER) ; do echo "#include \""$$f"\"" >> $@ ;
> done
> >>
> >> which puts the output of echo into a new file I sepcified.
> >>
> >> changed to this:
> >>
> >>        @for f in $(INCLUDE_DIR)/%.h); do \
> >>         echo "#include \""$$f"\""; \
> >>        done > '$@'
> >>
> >> which does the same thing but the writing is done after the loop
> >> completes
> >> (I think). Can someone please explain the difference and why it works?
> >
> > This is a question about shell programming, not about makefiles and is
> > probably better asked in help list for shell hacking.
> >
> > However, I can answer it for you :-)
> >
> > Note that the latter does NOT do all the writing after the loop
> > completes.  The file is written to, one line at a time, just as in the
> > first example.
> >
> > The first example says for each file, echo this one line and append it
> > (hence ">>" which is append) to the file address@hidden
> >
> > The second example says, for each file echo this one line, and redirect
> > all output from all commands run inside the for-loop to overwrite the
> > file $@ (hence ">" which is overwrite--i.e., truncate and write from
> > scratch).
> >
> > Some differences between the two:
> >
> >      1. If the file $@ exists before the command is invoked, then the
> >         first version will append the new content to the
> >         already-existing content... this is probably not what you want
> >         in this situation.  The second version will always recreate the
> >         file from scratch.
> >      2. If you wanted to print some content to stdout and some to the
> >         file, then you have to use the first version because the second
> >         version will put all stdout from all commands in the for loop
> >         into the file (well, you can play fancy tricks with file
> >         descriptors but it's hardly worth it).
> >
> >
> >
> > _______________________________________________
> > Help-make mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/help-make
> >
> >
>
> thanks to both of you. so the second version doesn't only write after the
> for loop is complete? i guess that's just how i am reading it since it is
> after the "done" keyword. what about this line tells the redirect to only
> print the one line above it: "done > '$@'" ?
>

The 'done' keyword says that the LOOP is done, not output. A loop can,
incidentally, be used as a scope for purposes of output redirection. These
examples might help make it clear:

address@hidden:~$ for i in a b c; do echo $i; done > foo
address@hidden:~$ cat foo
a
b
c

address@hidden:~$ { for i in a b c; do echo $i; done > foo; echo bar; } > bar
address@hidden:~$ cat foo
a
b
c
address@hidden:~$ cat bar
bar

But, as Paul said, this is really a Unix shell topic, and not Make. The
exact rules of the shell constructs depend on the shell you've told make to
use (defaullt=/bin/sh).

-- 
----- stephan beal
http://wanderinghorse.net/home/stephan/