Re: GUB fails to rebuild, configure issues?

2009-09-24 Thread Jan Nieuwenhuizen
Op dinsdag 11-08-2009 om 21:01 uur [tijdzone -0700], schreef Graham
Percival:
> On Tue, Aug 11, 2009 at 02:30:53PM +0200, John Mandereau wrote:
> > Le mardi 11 août 2009 à 04:46 -0700, Graham Percival a écrit :
> > > Odd behavior: if I rm -rf target/*/build/lilypond*,

> > > I thought that I'd seen error messages about needing to remove
> > > config.hh, but this last time that I cleaned things out and
> > > generated the error, there's no such message.
> 
> Got it now!
> 

> /home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh
> is out of date
>  *** Remove it and rerun autogen:
>  rm
> /home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh;
> ./autogen.sh

This is now fixed in master.  It turns out that git reset --hard
updates all time stamps, even on files that are not modified.

Unlike git checkout which makes a point of documenting the importance
of preserving/updating timestamps, git reset --hard blindly destroys
all timestamps on not-modified files.

This breaks configure-up-to-date checks, notably lilypond's.

Jan.

-- 
Jan Nieuwenhuizen  | GNU LilyPond - The music typesetter
Avatar®: http://AvatarAcademy.nl| http://lilypond.org



___
lilypond-devel mailing list
lilypond-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/lilypond-devel


Re: GUB fails to rebuild, configure issues?

2009-08-13 Thread John Mandereau
Le mardi 11 août 2009 à 21:01 -0700, Graham Percival a écrit :
> This changed the message, but the underlying problem is still
> there.  I tried another "make lilypond", without deleting
> anything.  (a few commits changed Documentation/general/ and css
> stuff, but nothing else happened on master)

I don't know what's happening exactly in GUB here (this is still a shame
I don't build it, I know).  Try to remove checksum file generated by
smart-autogen/smart-configure then call GUB again.  I hope a general fix
can be made in GUB in case it's worth it.

I'm afraid I must discard from now bug/failure reports on unclean builds
of Lily and docs so I can concentrate on switching to SCons.

Cheers,
John


signature.asc
Description: Ceci est une partie de message numériquement signée
___
lilypond-devel mailing list
lilypond-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/lilypond-devel


Re: GUB fails to rebuild, configure issues?

2009-08-11 Thread Graham Percival
On Tue, Aug 11, 2009 at 02:30:53PM +0200, John Mandereau wrote:
> Le mardi 11 août 2009 à 04:46 -0700, Graham Percival a écrit :
> > Odd behavior: if I rm -rf target/*/build/lilypond*,
> > target/*/src/lilypond*, target/*/status/lilypond*, and
> > target/*/installer/lilypond*   (I'm trying to be safe), it seems
> > to build ok.  But if I try to rebuild with "make lilypond", or if
> > I forget to clean one directory, it fails.
>
> This is certainly because GUB uses smart-autogen.sh, which runs
> autogen.sh if only if stepmake/aclocal MD5 sums have changed since last
> smart-autogen.sh invocation.
...
> I'm reverting the change I made.

This changed the message, but the underlying problem is still
there.  I tried another "make lilypond", without deleting
anything.  (a few commits changed Documentation/general/ and css
stuff, but nothing else happened on master)

Build failed for debian-ppc.

> > I thought that I'd seen error messages about needing to remove
> > config.hh, but this last time that I cleaned things out and
> > generated the error, there's no such message.

Got it now!

Overall console message:
...
 *** Stage: compile (lilypond, darwin-ppc)
Command barfed: cd 
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master
 && make -j1   TARGET_PYTHON="/usr/bin/env python"

Tail of target/linux-x86/log/build.log 
 rm 
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh;
 ./autogen.sh

make: *** 
[/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh]
 Error 1
Command barfed: cd 
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master
 && make -j1   TARGET_PYTHON="/usr/bin/env python"  
 Tail of target/linux-x86/log/build.log




target/linux-x86/log/build.log:

...
cat
/home/lilypond/gub/target/darwin-ppc/src/lilypond-git.sv.gnu.org--lilypond.git-master/GNUmakefile.in
>> GNUmakefile
chmod -w GNUmakefile
make[1]: Leaving directory
`/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master'

 ***
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh
is out of date
 *** Remove it and rerun autogen:
 rm
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh;
./autogen.sh

make: ***
[/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master/config.hh]
Error 1
Command barfed: cd
/home/lilypond/gub/target/darwin-ppc/build/lilypond-git.sv.gnu.org--lilypond.git-master
&& make -j1   TARGET_PYTHON="/usr/bin/env python"
Traceback (most recent call last):




___
lilypond-devel mailing list
lilypond-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/lilypond-devel


Re: GUB fails to rebuild, configure issues?

2009-08-11 Thread John Mandereau
Le mardi 11 août 2009 à 04:46 -0700, Graham Percival a écrit :
> Odd behavior: if I rm -rf target/*/build/lilypond*,
> target/*/src/lilypond*, target/*/status/lilypond*, and
> target/*/installer/lilypond*   (I'm trying to be safe), it seems
> to build ok.  But if I try to rebuild with "make lilypond", or if
> I forget to clean one directory, it fails.

> I thought that I'd seen error messages about needing to remove
> config.hh, but this last time that I cleaned things out and
> generated the error, there's no such message.

This is certainly because GUB uses smart-autogen.sh, which runs
autogen.sh if only if stepmake/aclocal MD5 sums have changed since last
smart-autogen.sh invocation.  The problem is, Werner requested that make
checks that configure script has a more recent modification time than
configure.in and stepmake/aclocal.m4, which may happen even if MD5 sums
of these files -- e.g. mtime of source files checked out from Git is the
time when the file was last checked out or updated after
merging/fast-forwarding/pulling.  I'm reverting the change I made.

To sum up, smart-autogen.sh is too smart for make, and this is one more
reason to switch the build system to SCons, which Jan almost convinced
me to do.

John


signature.asc
Description: Ceci est une partie de message numériquement signée
___
lilypond-devel mailing list
lilypond-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/lilypond-devel