svn commit: r348540 - head

[Enji Cooper]

> On Jun 3, 2019, at 12:06 PM, Rodney W. Grimes <freebsd at gndrsh.dnsmgr.net> wrote:
> 
>> Author: sobomax
>> Date: Mon Jun  3 15:34:00 2019
>> New Revision: 348540
>> URL: https://svnweb.freebsd.org/changeset/base/348540
>> 
>> Log:
>>  Rollback the rest of the botched r348521. Re-work would be posted to
>>  reviews. Sorry.
>> 
>>  Reported by:	Enji Cooper
>> 
>> Modified:
>>  head/Makefile.inc1
>> 
>> Modified: head/Makefile.inc1
>> ==============================================================================
>> --- head/Makefile.inc1	Mon Jun  3 15:32:42 2019	(r348539)
>> +++ head/Makefile.inc1	Mon Jun  3 15:34:00 2019	(r348540)
>> @@ -887,7 +887,7 @@ MTREEFLAGS+=	-W
>> INSTALLFLAGS+=	-h sha256
>> .endif
>> .if defined(DB_FROM_SRC) || defined(NO_ROOT)
>> -IMAKE_INSTALL=	INSTALL="${INSTALL} ${INSTALLFLAGS}"
>> +IMAKE_INSTALL=	INSTALL="install ${INSTALLFLAGS}"
> 
> This is a correct change,
> ngie please explain exactly why this is an exception to INSTALL?=install.

	There are edge cases I’ve had to work around in the build system in the past, where using $INSTALL would not work here, e.g., use “INSTALL=true”. This would break that use.
	There’s also absolutely no guarantee that “$INSTALL” knows how to interpret "$INSTALLFLAGS”; FreeBSD’s install(1) however, used during make distribution, installworld, installkernel, etc, does support these flags.
	The same goes for mtree as well; nmtree isn’t the same as the older version of mtree that’s on older (pre-9.x) systems (IIRC). It’s better to use known build/install tools to avoid maintenance/debugging burden.
Thanks,
-Enji