Re: VPATH?

[Harlan Stenn]

> >>> "Harlan" == Harlan Stenn <address@hidden> writes:
> 
>  Harlan> So in the past we have successfully created Makefile.am's that say:
>  Harlan> address@hidden@:@srcdir@/other/dir
> 
>  Harlan> foo_SOURCES= filefromdot filefromotherdir
> 
>  Harlan> and with 1.8 I am seein more places where this no longer works.
> 
>  Harlan> The documentation does not clearly state that
>  Harlan> augmenting VPATH is wrong.
> 
>  Harlan> Is it wrong to use an "augmented" VPATH with automake?
> 
> Right now, yes.  I don't know how it could have possibly worked
> in the past.  In fact I don't believe it ever worked, but maybe
> it depends on the definition we use for `to work' :)

The software tree compiled.

> For instance `make dist' has always assumed that files are
> either in the current directory, or in $(srcdir).  Therefore it
> would never distribute `filefromotherdir' [*].  I believe this
> apply to all Automake versions.

We never used the dist target.

> Could you give a more detailed example of what worked, when, and
> how?  "more places" is a bit fuzzy.

A subdir that had:

 Makefile.am:
 address@hidden@:@srcdir@/../common

 bin_PROGRAMS= foo

and no other files in it used to automatically get foo.c from
srcdir/../common/foo.c, and with automake-1.8 it no longer does.

I have 108 other Makefile.am's that are successfully doing this; every so
often another one will break and I have to move the "path" spec from VPATH
to the correct file names.  This happens Rarely.

> [*] I think this is not accurate if you limit yourself Solaris
>     or Tru64 Make.  But I'm assuming you don't.

This project Requires GNU make.

H