Re: [PATCH] automake: avoid harmful directory change before invoking valac

On 12/22/2014 01:36 PM, Yanko Kaneti wrote:
> On Mon, 2014-12-22 at 12:05 +0100, Stefano Lattarini wrote:
>> On 12/22/2014 11:32 AM, Yanko Kaneti wrote:
>>> On Fri, 2014-12-19 at 22:01 +0100, Stefano Lattarini wrote:
>>>> Hi, thanks for the patch, and sorry for the delay.
>>>>
>>>> On 12/02/2014 02:22 PM, Yanko Kaneti wrote:
>>>>> The current am__cd right before invoking valac invalidates
>>>>> any relative flags setup done before that.
>>>>>
>>>>> https://bugzilla.gnome.org/show_bug.cgi?id=740825
>>>>    >
>>>> BTW, this was already reported as http://debbugs.gnu.org/13002
>>>> So finally fixing it would likely be a good idea.
>>>>
>>>>> ---
>>>>>     bin/automake.in | 3 ++-
>>>>>     1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>
>>>> Such a fix requires at least a new test case and a NEWS entry.
>>>> Care to take at stab at writing them?
>>>
>>> Sorry,
>>> I looked a bit in the t/ directory but my eyes glazed over.
>>> Perhaps I am not the best person to attempt those.
>>>
>> It's OK if you can just post the snippets of relevant vala sources
>> and build options exposing the problem, and describe the steps to
>> reproduce such a problem.  I can synthesize a test case from there
>> myself then.
>
> It was exposed by a builddir != srcdir problem in gsound in gnome
> git https://git.gnome.org/browse/gsound/ (linked bugreport above)
> where a test program (gsound-play.vala) needs the gsound vapis which
> are also a build product.
>
> The way to reproduce it is:
>
> - autogen gsound with the workaround (4f8a78c) removed
> - mkdir _build & cd _build
> - ../configure
> - notice how the gsound-play compile fails  in the _srcdir_ (because
> of the the "am__cd")  but its arguments are trying to use ../gsound
> instead of the builddir gsound
>
> Hope that helps.
> Yanko
>
Since I know almost nothing about Vala, and I'm short of time these
days, I don't think I will go perusing into a third-party project
to find a test case; a minimal reproducer in form of Vala sources,
a Makefile.am fragment, and an high-level description of the expected
behavior) would be appreciated, and would help in speeding up the
fix.

Thanks,
   Stefano