On Fri, 24 Jun 2011 07:53:44 -0300, david@tethera.net wrote:
> From: David Bremner <bremner@debian.org>

Hi David,

Thanks for the fix. A couple of nit-picky comments on the commit itself
first:

> - c0961e6 introduced a missing slash and
> - cdf1c70a created a file and neglected to add it to clean

I'd put a little more detail in the commit message here:

- c0961e6 introduced a missing slash $(dir)$(LIBNAME)
- cdf1c70a created $(dir)/notmuch.h.gch and neglected to add it to clean

> The former seems to have been harmless, so maybe someone (Carl?) can
> check if $(dir)/$(LIBNAME) really needs to be removed?

This auxiliary text should be below the "---" line in the email so that
it won't be in the commit message after I do "git am". (As is, I'd have
to do extra work to "git commit --amend" this text away).

As for the actual question, yes, $(dir)/$(LIBNAME) is supposed to be
cleaned and is not getting cleaned. I've got libnotmuch.so.1.{1,2,3,4}.0
all sitting around never cleaned.

> -CLEAN := $(CLEAN) $(libnotmuch_modules) $(dir)/$(SONAME) $(dir)/$(LINKER_NAME) $(dir)$(LIBNAME) libnotmuch.a notmuch.aux notmuch.sym
> +CLEAN += $(libnotmuch_modules) $(dir)/$(SONAME) $(dir)/$(LINKER_NAME)
> +CLEAN += $(dir)/$(LIBNAME) libnotmuch.a notmuch.aux notmuch.sym 

Meanwhile, I notice now that "libnotmuch.a" is also wrong, (should be
"$(dir)/libnotmuch.a).

Clearly we could use some testing of our "make clean" target to ensure
it actually works. Does the Debian stuff not test anything like this?
(Maybe I'm thinking of typical testing done by autoconf/automake's "make
distcheck" that we haven't replicated yet.)

-Carl

-- 
carl.d.worth@intel.com