Re: -current broken when MAKEOBJDIRPREFIX is set (was: src is on NFS)

> On 20 Jul 2015, at 11:40, Simon J. Gerraty <sjg_at_juniper.net> wrote:
> 
> O'Connor, Daniel <darius_at_dons.net.au> wrote:
>>>> So, it seems MAKEOBJDIRPREFIX only works as an environmental variable
> 
>> Weird, I could have sworn I have set it on the command line and had it
>> work, but..
> 
> In most "normal" usage you will likely not notice a difference.
> It is only when a makefile is "being clever" that things go south when
> you break its expectations.

I guess, buildworld is pretty clever though :)

>>> I thought there was a check in src/Makefile for that.
>> 
>> Not so far as I can tell - it certainly gets quite far before blowing
>> up with a non useful error message :)
> 
> The check in src/Makefile is only guarding against MAKEOBJDIRPREFIX set
> in say make.conf, it explicitly discards the possibility of setting
> MAKEOBJDIRPREFIX on command line - which seems wrong.
> 
> Also with bmake, you *can* usefully set MAKEOBJDIRPREFIX in a makefile
> since the choice of .OBJDIR can be made after make starts reading
> makefiles.
> 
> So the current test is perhaps out dated on that score too - though
> only if something like auto.obj.mk is being used (-DWITH_AUTO_OBJ)
> which I don't think I've tested with buildworld.

I know sod all about *make but as an 'end user' a seat belt to warn about this problem would be Really Nice (tm).

--
Daniel O'Connor
"The nice thing about standards is that there
are so many of them to choose from."
 -- Andrew Tanenbaum
GPG Fingerprint - 5596 B766 97C0 0E94 4347 295E E593 DC20 7B3F CE8C