(toplevel patch) Configure in Makefile, version 3.

(toplevel patch) Configure in Makefile, version 3.

[Nathanael Nerode]

Forget the last one.  I came up with a better (temporarily, anyway)
 solution.  *sigh*

Index: Makefile.tpl
===================================================================
RCS file: /cvs/gcc/gcc/Makefile.tpl,v
retrieving revision 1.5.2.21
diff -u -r1.5.2.21 Makefile.tpl
--- Makefile.tpl	4 Dec 2002 01:41:48 -0000	1.5.2.21
+++ Makefile.tpl	5 Dec 2002 02:21:12 -0000
@@ -530,14 +530,8 @@
 
 GCC_FLAGS_TO_PASS = $(BASE_FLAGS_TO_PASS) $(EXTRA_GCC_FLAGS)
 
-# This is a list of the configure targets for all of the modules which
-# are compiled using the target tools.
-CONFIGURE_TARGET_MODULES =[+
-    FOR target_modules +] \
-	configure-target-[+module+][+
-    ENDFOR target_modules +]
-
-configure-target: $(CONFIGURE_TARGET_MODULES)
+configure-host: @configure_host_modules@
+configure-target: @configure_target_modules@
 
 # This is a list of the targets for which we can do a clean-{target}.
 CLEAN_MODULES =[+
@@ -566,6 +560,7 @@
 	@all_host_modules@ \
 	@all_target_modules@
 
+all-host: @all_host_modules@
 all-target: @all_target_modules@
 
 # Do a target for all the subdirectories.  A ``make do-X'' will do a
@@ -910,7 +905,7 @@
 
 .PHONY: all-[+module+] maybe-all-[+module+]
 maybe-all-[+module+]:
-all-[+module+]:
+all-[+module+]: configure-[+module+]
 	@r=`${PWD}`; export r; \
 	  s=`cd $(srcdir); ${PWD}`; export s; \
 	  $(SET_LIB_PATH) \
@@ -1139,16 +1134,19 @@
 	  $(HOST_CONFIGARGS) $${srcdiroption} \
 	  || exit 1
 
+# Don't remake gcc if it's already been made by 'bootstrap'; that causes
+# nothing but trouble.  This wart will be fixed eventually by moving
+# the bootstrap behavior to this file.
 .PHONY: all-gcc maybe-all-gcc
 maybe-all-gcc:
-all-gcc:
-	@if [ -f ./gcc/Makefile ] ; then \
+all-gcc: configure-gcc
+	@if [ -f gcc/xgcc ] ; then \
+	  exit 0 ; \
+	else \
 	  r=`${PWD}`; export r; \
 	  s=`cd $(srcdir); ${PWD}`; export s; \
 	  $(SET_LIB_PATH) \
 	  (cd gcc; $(MAKE) $(GCC_FLAGS_TO_PASS) all); \
-	else \
-	  true; \
 	fi
 
 # Building GCC uses some tools for rebuilding "source" files
@@ -1162,7 +1160,7 @@
 # in parallel.
 #
 .PHONY: bootstrap bootstrap-lean bootstrap2 bootstrap2-lean bootstrap3 bootstrap3-lean bootstrap4 bootstrap4-lean bubblestrap quickstrap cleanstrap restrap
-bootstrap bootstrap-lean bootstrap2 bootstrap2-lean bootstrap3 bootstrap3-lean bootstrap4 bootstrap4-lean bubblestrap quickstrap cleanstrap restrap: all-bootstrap
+bootstrap bootstrap-lean bootstrap2 bootstrap2-lean bootstrap3 bootstrap3-lean bootstrap4 bootstrap4-lean bubblestrap quickstrap cleanstrap restrap: all-bootstrap configure-gcc
 	@r=`${PWD}`; export r; \
 	s=`cd $(srcdir); ${PWD}`; export s; \
 	$(SET_LIB_PATH) \

Re: (toplevel patch) Configure in Makefile, version 3.

[Andrew Cagney]

Nathanael, `we have a problem'.

In addition to the last two evenings snapshot non-config's, I'm now seeing:

$ /home/cagney/PENDING/2002-12-04-init-frame-pc-return/src/configure 
--target=alpha-elf --enable-gdb-warnings=,-Werror
Configuring for a i686-pc-linux-gnu host.
Created "Makefile" in 
/home/cagney/PENDING/2002-12-04-init-frame-pc-return/carp/alpha-elf 
using "mt-frag"
$ ls
alpha-elf  config.status  Makefile  mt-frag

(Red Hat 7.2, i386)

Andrew

Re: (toplevel patch) Configure in Makefile, version 3.

[DJ Delorie]

> In addition to the last two evenings snapshot non-config's, I'm now seeing:
> 
> $ ls
> alpha-elf  config.status  Makefile  mt-frag

He he he.   Did you try typing "make" ?

Hopefully you'll be pleasantly surprised...

Re: (toplevel patch) Configure in Makefile, version 3.

[Andrew Cagney]

>> In addition to the last two evenings snapshot non-config's, I'm now seeing:
>> 
>> $ ls
>> alpha-elf  config.status  Makefile  mt-frag
> 
> 
> He he he.   Did you try typing "make" ?
> 
> Hopefully you'll be pleasantly surprised...

Interesting, wonder how much documentation this is going to affect.

Andrew

Re: (toplevel patch) Configure in Makefile, version 3.

[DJ Delorie]

> Interesting, wonder how much documentation this is going to affect.

That's a very good question.

Re: (toplevel patch) Configure in Makefile, version 3.

[Andrew Cagney]

DJ, humor me.


> He he he.   Did you try typing "make" ?
> 
> Hopefully you'll be pleasantly surprised...


So how does a GDB snapshot work again?  The documentation states:

$  gmake -f Makefile.in gdb.tar
Makefile.in:7776: *** missing separator.  Stop.

Andrew

Re: (toplevel patch) Configure in Makefile, version 3.

[Daniel Jacobowitz]

On Thu, Dec 05, 2002 at 03:30:29PM -0500, Andrew Cagney wrote:
> DJ, humor me.
> 
> 
> >He he he.   Did you try typing "make" ?
> >
> >Hopefully you'll be pleasantly surprised...
> 
> 
> So how does a GDB snapshot work again?  The documentation states:
> 
> $  gmake -f Makefile.in gdb.tar
> Makefile.in:7776: *** missing separator.  Stop.

Weren't you using src-release for this now?

However, src-release will need updating to actually run the configure
targets.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: (toplevel patch) Configure in Makefile, version 3.

[Andrew Cagney]

> On Thu, Dec 05, 2002 at 03:30:29PM -0500, Andrew Cagney wrote:
> 
>> DJ, humor me.
>> 
>> 
> 
>> >He he he.   Did you try typing "make" ?
>> >
>> >Hopefully you'll be pleasantly surprised...
> 
>> 
>> 
>> So how does a GDB snapshot work again?  The documentation states:
>> 
>> $  gmake -f Makefile.in gdb.tar
>> Makefile.in:7776: *** missing separator.  Stop.
> 
> 
> Weren't you using src-release for this now?
> 
> However, src-release will need updating to actually run the configure
> targets.

I'm doing what ever the doco tells me :-)
http://sources.redhat.com/gdb/current/onlinedocs/gdbint_15.html#SEC151

(hmm, the doco's out-of-date for the mainline - my bad, sorry :-()

back to the new problem:

Configuring utils...
loading cache ../config.cache
checking for a BSD compatible install... (cached) /usr/bin/install -c
checking host system type... sparc-sun-sunos4.1.1
checking target system type... sparc-sun-sunos4.1.1
checking build system type... sparc-sun-sunos4.1.1
creating ./config.status
creating Makefile
gmake configure-target \
     ALL_GCC="" ALL_GCC_C="" ALL_GCC_CXX="" \
     CC_FOR_TARGET="cc" CXX_FOR_TARGET="g++"
gmake[3]: Entering directory `/notnfs/cagney/GDB/src'
Makefile:6618: warning: overriding commands for target 
`sun4/libiberty/Makefile'
Makefile:1258: warning: ignoring old commands for target 
`sun4/libiberty/Makefil
e'
gmake[3]: *** No rule to make target `cc', needed by 
`sun4/libstdc++-v3/multilib
.out'.  Stop.
gmake[3]: Leaving directory `/notnfs/cagney/GDB/src'
gmake[2]: *** [do-proto-toplev] Error 2
gmake[2]: Leaving directory `/notnfs/cagney/GDB/src'
gmake[1]: *** [gdb-tar] Error 2
gmake[1]: Leaving directory `/notnfs/cagney/GDB/src'
gmake: *** [gdb.tar] Error 2
cagney@torrens$

It's pretty easy to reproduce.

	gmake -f src-release gdb.tar

Andrew

Re: (toplevel patch) Configure in Makefile, version 3.

[DJ Delorie]

Right, Nate already promised a patch for this sometime today.

> gmake[3]: *** No rule to make target `cc', needed by 
> `sun4/libstdc++-v3/multilib.out'.  Stop.

Re: (toplevel patch) Configure in Makefile, version 3.

[Andrew Cagney]

>> In addition to the last two evenings snapshot non-config's, I'm now seeing:
>> 
>> $ ls
>> alpha-elf  config.status  Makefile  mt-frag
> 
> 
> He he he.   Did you try typing "make" ?
> 
> Hopefully you'll be pleasantly surprised...

Ok, now that I've recovered from my `pleasant suprise' :-) I'm finding 
that, given a build that barfs in GDB, it tries to re-configure BFD when 
re-made.

It's being run with `gmake -k -j 2'.  Anyone else noticed this?

Andrew

Re: (toplevel patch) Configure in Makefile, version 3.

[Nathanael Nerode]

> Ok, now that I've recovered from my `pleasant suprise' :-) I'm finding 
>that, given a build that barfs in GDB, it tries to re-configure BFD 
>when 
>re-made.
>
> It's being run with `gmake -k -j 2'. Anyone else noticed this?

Not me.

Have you got it working when the build doesn't crash in gdb?...
I've done oodles of builds with the current versions of the files, but 
I'm sure there are builds I haven't tried. :-/

The only reasons it should try to reconfigure BFD are:
* bfd/Makefile was changed (or 'touch'ed) by something
* libiberty/Makefile was changed (or 'touch'ed) by something
* config.status at the top level was changed (or 'touch'ed) by something

I can't imagine why either of those things would happen.

Is it really a problem, anyway?  BFD is reconfigured with the same 
arguments it was configured with the previous time...

It *will* try to rebuild BFD.  That is correct.  BFD's Makefile should 
be clever enough to do nothing in this case, though it might not be.

Hmm.  When GNU make runs BFD's Makefile, maybe BFD's Makefile decides 
that it's out of date?  I haven't poked BFD's Makefile enough.

Can you figure out which level of Make is trying to rebuild BFD's 
Makefile?

--Nathanael