Re: [Automake-NG] [PATCH 15/17] [ng] check: accept dot-less entries in $(TEST_EXTENSIONS)

[Stefano Lattarini]

On 05/22/2012 11:35 PM, Akim Demaille wrote:
> 
> Le 22 mai 2012 à 22:50, Stefano Lattarini a écrit :
> 
>> Since the beginning of the parallel-tests harness, we've been requiring
>> that the entries in $(TEST_EXTENSIONS) begin with a dot (as suggested by
>> the name "extensions").  But it's clear that if an user writes something
>> like:
>>
>>    TEST_EXTENSIONS = test sh
>>
>> he actually means:
>>
>>    TEST_EXTENSIONS = .test .sh
>>
>> So let's try to be more user-friendly, and some DWIM processing on the
>> content of $(TEST_EXTENSIONS).
> 
> Well, I'm not so sure about this one.  What if someone wants to
> write tests with ,ext?
>
That is definitely asking for trouble IMHO.  I'm not sure I want us to commit
to support such an usage, until there is any real user request for it.

Let's take a step back for the moment: drop this patch, and replace it with one
that introduces a make runtime check verifying that all $(TEST_EXTENSIONS)
entries end with a dot, erroring out if this is not the case.  This is just
a minimal change w.r.t. the behaviour of mailine Automake, and keep us free to
define a proper semantic for dot-less extension in a second time, should the
need ever arise.

OK?

> Reading the diff I'm not sure it wouldn't
> work, yet we'd have two different ways to say the same thing, which
> I don't think is a good property.
> 
> Btw, it would be nice if Automake-ng could use a single convention
> where automake uses several.
>
>      AM_DEFAULT_SOURCE_EXT = .cpp
> 
> Also, OBJEXT does not include the dot, whereas EXEEXT does.
>
That's because $(OBJEXT) is always non-empty ('.o' or '.obj'), while
$(EXEEXT) might be empty (it is on most real POSIX hosts, while it is
'.exe' on Cygwin and minGW).

Anyway, I think the assumption "$(OBJEXT) is always dot-less" is so
entrenched in the current codebase that trying to change them would
cause endless troubles (more than it's worth).

> With %-pattern rules, everything could be made alike, no?
>

Regards,
  Stefano