Re: [LTP] [PATCH 2/3] Make system enhancements, Draft 1

[Garrett C. <yan...@gm...>]

On Wed, Jul 8, 2009 at 4:24 PM, Mike Frysinger<va...@ge...> wrote:
> On Monday 06 July 2009 19:02:50 Garrett Cooper wrote:
>> 1. Write up a document on how to use the new Make system
>> (README.mk-devel), so others may assist in the work that we're
>> spearheading at Cisco and contributing back to LTP, and we can begin
>> dropping the adhoc Makefiles.
>> 2. Add master_rules.mk which contains canned rules for environment
>> setup, leaf directories, and trunk directories to reduce Make logic
>> noise in the existing LTP Makefile's. This is a more condensed and
>> straightforward version of the original master_rules.mk file
>> contributed and never integrated into LTP in the past.
>> 3. Add master_include.mk, a file which stands as the defacto include
>> Makefile for all . This was created out of the comment provided by
>> Mike Frysinger w.r.t. a master include file.
>>
>> Note: This patch has been largely tested, but some of the additional
>> functionality and rework in generic_trunk_target needs to be tested
>> more prior to commit; hence it's a Draft 1 patch.
>
> "srcdir" should refer to the current directory's source tree.  "top_srcdir"
> should refer to the top level source tree.  building in-tree means srcdir will
> be ".".  this is how autotools works and allows for easy out-of-tree
> compilation.  same goes for "builddir" and "top_builddir", not this "objxxx"
> convention.

Ah, ok. Will do -- thanks!

> for the toolchain, do you purposefully only document CC and CFLAGS ?

Where did I do that?

> i'm not familiar with this "MAKEOPTS=';'" thing you refer to.  what's that all
> about ?

       -r, --no-builtin-rules
            Eliminate use of the built-in implicit rules.  Also clear
out the default list of suffixes for suffix rules.

There isn't a way to disable -r in make-3.81 except by removing it
from MAKEOPTS (that's how it was enabled in our make system at Cisco
in my group). `MAKEOPTS=;' is the quick way out. Nasty PITA...

> you list the copyright info twice

Please see my other comment.

> master_rules.mk shouldnt be a '-include' ... we want an error if that doesnt
> exist

I only did that to ensure that this commit could go in, irrespective
of PATCH 3/4, but I agree, it should be enabled once both changes are
committed.

> your %.a rule is missing a call to ranlib on the archive

Ok. How do I do that call exactly?

ar [blah]
ranlib [blah]

maybe?

Thanks!
-Garrett