Re: [PATCH 1/2] Build symbol-test with make instead of hardcoding in symbol-hiding.

Subject: Re: [PATCH 1/2] Build symbol-test with make instead of hardcoding in symbol-hiding.

Date: Fri, 25 Nov 2011 16:12:24 +0100

To: David Bremner

Cc: notmuch

From: Amadeusz Żołnowski


Excerpts from David Bremner's message of 2011-11-25 14:04:44 +0100:
> On Fri, 25 Nov 2011 11:54:51 +0100, Amadeusz Żołnowski <aidecoe@aidecoe.name> wrote:
> >  
> > +$(dir)/symbol-test: $(dir)/symbol-test.o
> > +    $(call quiet,CC) $^ -o $@ -Llib -lnotmuch -lxapian
> 
> I'm a bit surprised no -I is neaded here. Is that taken care in the CC
> call?

-I is for headers which are not used when linking.  symbol-test.o is
compiled according to rule defined earlier in some parent Makefile.


> >  .PHONY: test check
> > -test:    all $(dir)/smtp-dummy
> > +test:    all $(dir)/smtp-dummy $(dir)/symbol-test
> >      @${dir}/notmuch-test $(OPTIONS)
> 
> The instructions were already wrong, but maybe they should be updated to
> point out that calling the test scripts directly requires smtp-dummy and
> symbol-test already be built. Or maybe the tests should check for these
> binaries as pre-reqs along with emacs, gpg, etc... 
> 
> > +       -e "/^(emacs.expected-output|smtp-dummy|smtp-dummy.c|test-verbose|symbol-test.cc|symbol-test|symbol-test.o)/d" \
> 
> I wonder if this should be shortened (and made a bit sloppier) with the
> use of regexes.

Thanks it's so straightforward.  I found this line quickly without deep
analysis of all test-framework. :-)  It could only be made more flexible
- put list of this files into some separate text file.  But that's
subject for another patch.  Purpose of this one is to make building out
of tree working.


--
Amadeusz Żołnowski
signature.asc (application/pgp-signature)

Thread: