Re: [PATCH v3] new option: object-shortname

[Mathieu Lirzin]

Thomas Martitz <address@hidden> writes:

> Am 02.03.2017 um 18:47 schrieb Mathieu Lirzin:
>> Hello Thomas,
>>
>> Thomas Martitz <address@hidden> writes:
>>
>>> Am 16.01.2017 um 13:56 schrieb Thomas Martitz:Hello Jim and other folks,
>>>>
>>>> This is a friendly reminder. I've added a test suite to the
>>>> patch. Please review.
>>>>
>>>> Thank you and best regards,
>>>> Thomas Martitz
>>>>
>>>
>>> this is one more friendly reminder. Please review my patch set.
>>
>> As the newly appointed Automake maintainer, I take up the task of
>> reviewing your patch.
>>
>
> Thank you, this is highly appreciated. I hope we can resolve remaining
> issues.
>
> Also, I'm very happy to learn that Automake has a new maintainer. Hope
> you'll like it!
>
>> Currently the Automake manual recommends against using '_SHORTNAME'
>> unless required for portability, see:
>>
>>    ‘maude_SHORTNAME’
>>         On some platforms the allowable file names are very short.  In
>>         order to support these systems and per-target compilation flags at
>>         the same time, Automake allows you to set a “short name” that will
>>         influence how intermediate object files are named.
>>    [...]
>>         This facility is rarely needed in practice, and we recommend
>>         avoiding it until you find it is required.
>>
>> As a consequence, I am a bit reluctant to add this new
>> 'object-shortname' option which seems like a convenience for something
>> not recommended.
>
> I can understand your argument, however my patch is about *avoiding*
> to use _SHORTNAME, because explicitely writing out _SHORTNAME
> everywhere is a nightmare and doesn't work in all circumstances
> (inside conditionals). The option also does not promote the use of
> _SHORTNAME in any way.
>
> My patch isn't about convenience. The main purpose is to solve the
> second issue while also avoiding the use of _SHORTNAME every where as
> a side effect.
>
> For my use case I *have to* adjust the file name of the generated
> objects because otherwise the same source file compiles to different
> objects, depending on where make is invoked. This is currently
> incorrect behavior and I'm just trying to make my use case work.

Hum I fail to properly understand this issue.  It would be great if you
could provide an example with instructions to reproduce it.

> As an additional note, the option should be combined with
> subdir-objects, And as a result the object files are placed next to
> the source files with the same file name, only prefixed by the target
> name (and not the directory structure between the working dir of the
> make invocation and the source file).
>
> I think the manual does not recommend using it simply because it's not
> needed most of the time.
>
>>
>> Having said that, not being able to use '_SHORTNAME' with %reldir% and
>> %canon_reldir% inside Makefile snippets seems like a bug.  What about
>> trying to fix it without the 'object-shortname' option?
>>
>
> _SHORTNAME works with %reldir%. The problem is doesn't work inside
> conditionals, including when a Makefile fragment (that uses %reldir%)
> is conditionally included.
>
>
>> I hope my feedback after such long review delay will not be too
>> frustrating.  :/
>>
>
> Maybe I have to explain my use case better? Is it not clear to you?

Yes this is not clear to me. :)

A concrete example will hopefully help.

Thanks.

-- 
Mathieu Lirzin
GPG: F2A3 8D7E EB2B 6640 5761  070D 0ADE E100 9460 4D37