Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160117223247.GK238@brightrain.aerifal.cx>
Date: Sun, 17 Jan 2016 17:32:48 -0500
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: Support for out-of-tree build

On Thu, Dec 10, 2015 at 01:39:44AM +0000, Petr Hosek wrote:
> On Tue, Dec 8, 2015 at 4:39 PM Rich Felker <dalias@...c.org> wrote:
> 
> > I get:
> >
> > make: *** No rule to make target `obj/include/bits', needed by
> > `obj/include/bits/alltypes.h'.  Stop.
> >
> > I think this is because $(dir...) intentionally leaves the trailing
> > slash, so you have to remove it to get the desired behavior.
> >
> 
> I'm not sure why I don't see the same behavior in my setting, but I used
> make -p and you're correct, $(dir ...) leaves the trailing slashes.
> 
> 
> > First I tried:
> >
> > OBJ_DIRS = $(sort $(patsubst %/,%,$(dir $(ALL_LIBS) $(ALL_TOOLS) $(OBJS)
> > $(GENH) $(GENH_INT))) $(addprefix obj/, crt crt/$(ARCH) include))
> >
> > but this does not seem to be sufficient, because of:
> >
> > $(OBJS) $(LOBJS): | $(sort $(dir $(OBJS)))
> >
> > Adding patsubst there would work too, but I wonder if it even makes
> > sense to have fine-grained dir dependencies, rather than just making
> > all the targets in this section of the makefile depend on $(OBJ_DIRS).
> > Thoughts?
> >
> 
> I've used the approach you suggested yesterday over IRC which simplifies
> the Makefile a bit.
> 
> 
> > ../configure: Makefile already exists in the working directory
> >
> > This isn't what I would expect either. Normally you want to be able to
> > re-run configure with different options. There's probably a little bit
> > of subtlety here because we would like to avoid overwriting an actual
> > file (if somebody copied and edited a makefile here) but happily
> > replace a symlink.
> >
> 
> I've expanded the check to cover both the existing file and symlink case.
> 
> 
> > Aside from these issues it seems to be okay/working for me. I'd
> > appreciate comments from anyone else who's tried it.
> >
> 
> I've been using the previous version in my setup for about two weeks now
> and haven't found any further issue but I'd like to hear from others if
> they found any issues.

Hi again Petr!

Sorry it's taken so long to get around to actually committing this.
I've had a lot keeping me busy aside from musl, and was waiting both
to get some feedback from other users and to have a good chance to try
it out in real-world usage myself.

The good news is that everything looks fine. I've got my own
foreign-arch builds all converted over to build from my main source
tree, which is going to greatly reduce my frustration of having
in-progress cruft (that was later finished elsewhere) sitting around
in different cloned trees.

As such I've committed/pushed your patch. If anything else turns up
that should be fixed/improved that can be done later but I'm quite
happy with what we've got now. Thanks for all your work on it and your
patience with the process.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.