Re: CVS commit: src/usr.bin/make

[Ryo ONODERA <ryo_on%yk.rim.or.jp@localhost>]

From: christos%astron.com@localhost (Christos Zoulas), Date: Wed, 10 Jul 2013 
15:08:39 +0000 (UTC)

> In article <20130710.232445.202823964487691297.ryo_on%yk.rim.or.jp@localhost>,
> Ryo ONODERA  <ryo_on%yk.rim.or.jp@localhost> wrote:
>>From: christos%astron.com@localhost (Christos Zoulas), Date: Wed, 10 Jul 2013
>>13:08:49 +0000 (UTC)
>>
>>> In article <krbu04$2j1$1%ger.gmane.org@localhost>,
>>> Christos Zoulas <christos%astron.com@localhost> wrote:
>>>>>Module Name:       src
>>>>>Committed By:      sjg
>>>>>Date:              Sat Jul  6 18:19:17 UTC 2013
>>>>>
>>>>>Modified Files:
>>>>>   src/usr.bin/make: main.c var.c
>>>>>
>>>>>Log Message:
>>>>>If using gmake's MAKELEVEL; use it the same way
>>>>
>>>>Now you put it back the way it was before which is wrong. Gmake does not
>>>>behave this way. Before your change the following Makefile printed:
>>> 
>>> This is still broken. Renaming the variable is not a fix either.
>>> Consider the case where we switched from bmake to gmake as the
>>> pkgsrc wrapper. The packages that depend on $MAKELEVEL being 0 or
>>> unset would still break. So the proper fix for the 2 packages that
>>> broke is to make pkgsrc unset MAKELEVEL before invoking gmake.
>>> 
>>> So to fix those:
>>> - revert the changes so that MAKELEVEL again works the same way as in gmake.
>>> - add glue to invoke gmake with MAKELEVEL unset.
>>
>>Hi,
>>
>>I feel second idea is not so good.
>>Because pkgsrc creates symlink, work/.tools/bin/make, for our make and gmake.
>>So with gmake in USE_TOOLS, it is symlink for /usr/pkg/bin/gmake,
>>and without gmake in USE_TOOLS, it is for /usr/bin/make.
>>I cannot distinguish them with filename.
> 
> It is simple:
> 
> 1. You unset it for all (easiest+safest)
> 2. You create a shell script instead of a symlink that does:
>     unset MAKELEVEL && exec /usr/pkg/bin/gmake "$@"

Thanks for your explanation.

In pkgsrc case, shell script may work.
Outside pkgsrc, some program may be broken by this make(1) change.
I vote to "revert the changes so that MAKELEVEL again works
the same way as in gmake".

Thank you.

--
Ryo ONODERA // ryo_on%yk.rim.or.jp@localhost
PGP fingerprint = 82A2 DC91 76E0 A10A 8ABB  FD1B F404 27FA C7D1 15F3