Texino build problem

Texino build problem

[Mark Mitchell]

On mips-sgi-irix6.4 I now get:

  make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
  make[3]: *** No rule to make target `-c', needed by `all'.  Stop.

Indeed, it looks like:

  all: Makefile $(LIBRARIES) $(DATA)

  DATA =  $(INSTALL_DATA)
  INSTALL_DATA = ${INSTALL} -m 644
  INSTALL = ../../../texinfo/../install-sh -c

Does anyone comprehend this situation?

-- 
Mark Mitchell <mmitchell@usa.net>
http://home.earthlink.net/~mbmitchell
Consulting Services Available

Re: Texino build problem

[H.J. Lu]

> 
> 
> On mips-sgi-irix6.4 I now get:
> 
>   make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
>   make[3]: *** No rule to make target `-c', needed by `all'.  Stop.
> 
> Indeed, it looks like:
> 
>   all: Makefile $(LIBRARIES) $(DATA)
> 
>   DATA =  $(INSTALL_DATA)
>   INSTALL_DATA = ${INSTALL} -m 644
>   INSTALL = ../../../texinfo/../install-sh -c
> 
> Does anyone comprehend this situation?
> 

I got the same problem. The rebuilt Makefile.in files under texinfo
all look very strange.

-- 
H.J. Lu (hjl@gnu.org)

Re: Texino build problem

[Manfred Hollstein]

On Thu, 26 March 1998, 17:16:47, mmitchell@usa.net wrote:

    
    On mips-sgi-irix6.4 I now get:
    
      make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
      make[3]: *** No rule to make target `-c', needed by `all'.  Stop.
    
    Indeed, it looks like:
    
      all: Makefile $(LIBRARIES) $(DATA)
    
      DATA =  $(INSTALL_DATA)
      INSTALL_DATA = ${INSTALL} -m 644
      INSTALL = ../../../texinfo/../install-sh -c
    
    Does anyone comprehend this situation?

Yep, I do. Makefile.in is now a generated file which depends on

  Makefile.in -> { Makefile.am, configure.in, aclocal.m4}

We now have to remember, that we need to rebuild the various
Makefile.in files after changing configure.in (this unfortunately
needs automake-1.2f, which I  don't know where to get from. Jeff, do
you know?). You obviously are having automake-1.2e installed (as I do,
too), which generates these broken Makefile.in's.

The current sources in CVS look consistent with only timestamp
problems. The following command should do the trick (it did it for me):

$ cd egcs/texinfo
$ touch `find . -name Makefile.in -print`

If configure is out of date, you need to execute the following
commands in between the two:

$ mv configure configure.orig
$ autoconf
$ touch --file=configure.orig configure
$ autoheader
$ touch --file=configure config.h.in

manfred

Re: Texino build problem

[Richard Henderson]

On Thu, Mar 26, 1998 at 05:16:47PM -0800, Mark Mitchell wrote:
> 
> On mips-sgi-irix6.4 I now get:
> 
>   make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
>   make[3]: *** No rule to make target `-c', needed by `all'.  Stop.
> 
> Indeed, it looks like:
> 
>   all: Makefile $(LIBRARIES) $(DATA)
> 
>   DATA =  $(INSTALL_DATA)
>   INSTALL_DATA = ${INSTALL} -m 644
>   INSTALL = ../../../texinfo/../install-sh -c
> 
> Does anyone comprehend this situation?

It appears to be a bug in automake.   I hacked around it with

Index: automake.in
===================================================================
RCS file: /cvs/cvsfiles/devo/automake/automake.in,v
retrieving revision 1.20
diff -c -p -d -r1.20 automake.in
*** automake.in 1998/02/15 18:55:27     1.20
--- automake.in 1998/04/02 11:38:57
*************** sub am_install_var
*** 5816,5821 ****
--- 5816,5824 ----
            # all, or install targets.
            next if $X eq 'EXTRA';
  
+           # Don't see INSTALL_DATA wrongly.
+           next if $X eq 'INSTALL';
+ 
            # A blatant hack: we rewrite each _PROGRAMS primary to
            # include EXEEXT when in Cygwin32 mode.
            if ($seen_exeext && $primary eq 'PROGRAMS')

Perhaps someone else can comment on a proper solution?


r~

Re: Texino build problem

[Raja R Harinath]

Manfred Hollstein <manfred@s-direktnet.de> writes:
> On Thu, 26 March 1998, 17:16:47, mmitchell@usa.net wrote:
>     On mips-sgi-irix6.4 I now get:
>     
>       make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
>       make[3]: *** No rule to make target `-c', needed by `all'.  Stop.
>     
>     Indeed, it looks like:
>     
>       all: Makefile $(LIBRARIES) $(DATA)
>     
>       DATA =  $(INSTALL_DATA)
>       INSTALL_DATA = ${INSTALL} -m 644
>       INSTALL = ../../../texinfo/../install-sh -c
>     
>     Does anyone comprehend this situation?
> 
> Yep, I do. Makefile.in is now a generated file which depends on
> 
>   Makefile.in -> { Makefile.am, configure.in, aclocal.m4}
> 
> We now have to remember, that we need to rebuild the various
> Makefile.in files after changing configure.in (this unfortunately
> needs automake-1.2f, which I  don't know where to get from. Jeff, do
> you know?). You obviously are having automake-1.2e installed (as I do,
> too), which generates these broken Makefile.in's.

Nope.  Automake 1.2h isn't of any help, and unless things are very
different, Automake 1.2f wouldn't be either.  The problem caused by the
replacement of `AC_PROG_INSTALL' with `EGCS_PROG_INSTALL'.  `automake'
knows `AC_PROG_INSTALL' defines a special var called `INSTALL_DATA', but
it doesn't know that EGCS_PROG_INSTALL does too.  `automake' also
matches any *_DATA variables, and generates a Makefile that tries to
install the files listed in those variables.

One fix would be (I haven't tested this) to either rename
EGCS_PROG_INSTALL in acinclude.m4 to AC_PROG_INSTALL, or say

	AC_DEFINE([AC_PROG_INSTALL],[EGCS_PROG_INSTALL])

in acinclude.m4.

BTW, automake-1.2[fh] can be found at ftp://ftp.cygnus.com/pub/home/tromey .
 
> The current sources in CVS look consistent with only timestamp
> problems. The following command should do the trick (it did it for me):
> 
> $ cd egcs/texinfo
> $ touch `find . -name Makefile.in -print`
> 
> If configure is out of date, you need to execute the following
> commands in between the two:
> 
> $ mv configure configure.orig
> $ autoconf
> $ touch --file=configure.orig configure
> $ autoheader
> $ touch --file=configure config.h.in

That should work.

- Hari
-- 
Raja R Harinath ------------------------------ harinath@cs.umn.edu
"When all else fails, read the instructions."      -- Cahn's Axiom
"Our policy is, when in doubt, do the right thing."   -- Roy L Ash

Re: Texino build problem

[Tom Tromey]

>> make[3]: Entering directory `/pooma/mitchell/egcs/objdir/texinfo/lib'
>> make[3]: *** No rule to make target `-c', needed by `all'.  Stop.

>> all: Makefile $(LIBRARIES) $(DATA)
>> DATA =  $(INSTALL_DATA)
>> INSTALL_DATA = ${INSTALL} -m 644
>> INSTALL = ../../../texinfo/../install-sh -c

Raja> One fix would be (I haven't tested this) to either rename
Raja> EGCS_PROG_INSTALL in acinclude.m4 to AC_PROG_INSTALL, or say
Raja> 	AC_DEFINE([AC_PROG_INSTALL],[EGCS_PROG_INSTALL])
Raja> in acinclude.m4.


Ian Taylor forwarded me this message.  Suspected automake bugs should
be sent to <automake-bugs@gnu.org>, since I don't read the egcs list.

The proposed fix won't work.

The story is: automake magically knows about INSTALL_DATA, and doesn't
expect to see it defined by configure.  If it is, then automake gets
confused and thinks that `INSTALL_DATA' is a user-defined `_DATA'
variable.  Even if EGCS_PROG_INSTALL is renamed, automake will
probably still discover the `AC_SUBST(INSTALL_DATA)' -- which is what
is really causing the problem.


Could somebody tell me why EGCS_PROG_INSTALL is needed?  What does it
do?


The appended patch should fix the problem.  I added it and a test case
to my automake sources.

Tom

Index: automake.in
===================================================================
RCS file: /home/tromey/gnu/CVSROOT/automake/automake.in,v
retrieving revision 1.555
diff -u -r1.555 automake.in
--- automake.in	1998/04/01 04:10:39	1.555
+++ automake.in	1998/04/03 20:08:16
@@ -4273,7 +4273,8 @@
 	# FIXME.
 	if (/$AC_SUBST_PATTERN/o
 	    && $1 ne 'ANSI2KNR'
-	    && $1 ne 'INSTALL_SCRIPT')
+	    && $1 ne 'INSTALL_SCRIPT'
+	    && $1 ne 'INSTALL_DATA')
 	{
 	    $configure_vars{$1} = 1;
 	}

Re: Texino build problem

[Jeffrey A Law]

  In message < m11zvepste.fsf@creche.cygnus.com >you write:
  > Could somebody tell me why EGCS_PROG_INSTALL is needed?  What does it
  > do?
To avoid losing behavior with the default PROG_INSTALL -- it will
select an install program (I don't remember which one) which does
not work correctly.

jeff

Re: Texino build problem

[Jim Wilson]

	To avoid losing behavior with the default PROG_INSTALL -- it will
	select an install program (I don't remember which one) which does
	not work correctly.

I believe the problematic program was the `installbsd' program.  It looks like
the Cygnus autoconf sources get this right because of Jeff Law's Sept 4
patch, but I don't know if there are any released autoconf versions that
handle this correctly yet.

Jim

Re: Texino build problem

[Tom Tromey]

Jeff> At the time there was not autoconf maintainer; now that there is
Jeff> we should probably send that patch (and other local patches) to
Jeff> him.

I already sent him all the autoconf changes we made at Cygnus.
He said he'd probably incorporate them without change.

I don't know how far he is from doing a release.

Tom

Re: Texino build problem

[Jeffrey A Law]

  In message < 199804070130.SAA11301@rtl.cygnus.com >you write:
  > 	To avoid losing behavior with the default PROG_INSTALL -- it will
  > 	select an install program (I don't remember which one) which does
  > 	not work correctly.
  > 
  > I believe the problematic program was the `installbsd' program.  It looks like
  > the Cygnus autoconf sources get this right because of Jeff Law's Sept 4
  > patch, but I don't know if there are any released autoconf versions that
  > handle this correctly yet.
That sounds correct (installbsd).  And yes, I tweaked our version of
autoconf to avoid installbsd.

At the time there was not autoconf maintainer; now that there is we
should probably send that patch (and other local patches) to him.
jeff