Subject: Re: Problem in new toolchain builds (need comments)
To: Simon J. Gerraty <>
From: James Chacon <>
List: tech-toolchain
Date: 10/25/2001 04:13:10
>> 1. Where is all this distinction documented? I can't fathom any of this from
>> the man page. (nor the source especially from a cursory glance)
>Its in the source.  The man page attempts to explain some of this but
>since the man page is written by folk who typically wrote the source
>they often assume too much is already understood.  I'm as guilty as
>Much of the distinction between *MAKEFLAGS MFLAGS and MAKE is historic
>and is only mentioned briefly at a couple of points within the source
>and not at all in the man page.  Familiarity with other makes is
>perhaps part of the assumptions mentioned above.

Fair enough. I think I have it now based on some experimenting also.

>> 2. It doesn't work as one would expect. If you add the vars to any of either
>> MAKEFLAGS, .MAKEFLAGS, .MAKEOVERRIDES, etc they all get passed to the followon
>> makes. But they don't seem to get parsed apart again into var=value in the
>> sub-makes.
>I can only suggest you are doing something wrong:

I was....*whap*

>it works the same with the two alternatives you see commented out too.
>BTW I'm assuming your make(1) is reasonably up to date as .MAKEOVERRIDES
>and POSIX compatible export of command line vars was only implemented
>within the last 6 months. 

It was back from Feb...Ok, that might explain my problems...I'm usually
better at catching that...

>If I remove the .ifndef, then yes the shell code gets executed for
>each sub-make, but that's why the .ifndef is there.

Which var is the preference? I'm using .MAKEFLAGS since that seems to be
the one that the docs imply is where new env vars go you want passed along.