Re: Using Emacs with Autotools projects : separate source and binaries directories

[Bob Proulx]

fuhz.fr@gmail.com wrote:
> In my opinion, Emacs and automake are quite mature piece of software.
> The problem may be elsewhere.

At this point I can't guess where the problem would be.

> On the first compilation, there isn't any problem, hello-2.3 is
> definitely awesome.

:-)

> >   $ echo '#error forced error here' >> ../hello-2.3/src/hello.c
> > Then compile inside emacs again.
> >   M-x compile
> 
> When you are calling 'M-x compile' what is your working directory ?
> When I do this, I am in hello-2.3-mybuild.

I am in the same directory.  This is visible in the make output as the
last directory listed.  (That is, "/var/tmp/hello-2.3-mybuild".)

> > In emacs running C-x ` takes me to the next error.
> 
> I have exactly the same output, but Emacs do not find the source code.

Hmm...

> So, he asks me : "Where is your source directory ?"
> 
> The next message is shown, asking me for the location of the source
> directory :
> 'Find this error in (default ../../hello-2.3/src/hello.c): ~/mailing/
> hello-2.3-mybuild/'
> 
> And that's the point. The specified path '../../hello-2.3/src/hello.c'
> is not correct.
> Because it is relative to src directory.
> It should be '../hello-2.3/src/hello.c'.

Uhm... But in my output it also shows ../../hello-2.3/src/hello.c the
same as mine here does and yet my emacs finds it okay.  Emacs tracks
the compilation directory by tracking the "Entering directory" and
"Leaving directory" messages from make.  Therefore from the src
directory the path is two up and over but emacs tracks that okay.

Perhaps the make options have been modified for you such that make is
no longer producing directory tracking information?

> When I compile in src directory, the relative path is correctly set.
> And there is not any problem to find the source.
> 
> I you are doing the same, it seems our Emacs have different behaviour.

So it would seem.

> What I was trying to say is that VPATH builds seems to be less
> efficient.

Hmm... Measurably different?  Something that would show up in a
benchmark?  I am very suspicious of subjective feelings here.

Bob