tree-ssa branch: Fortran libraries don't build

tree-ssa branch: Fortran libraries don't build

[Joe Buck]

When trying to bootstrap the tree-ssa branch on a Red Hat 7.2 i686 system,
the build of libI77 fails in the configure step.  The top-level configure
line is

/u/jbuck/gnu/src/cvs/tree-ssa/gcc/configure \
    --prefix /usr/localdisk/gcc-cvs/tree-ssa \
    --with-as=/usr/localdisk/binutils-2.13.1/bin/as \
    --with-ld=/usr/localdisk/binutils-2.13.1/bin/ld \
    --enable-languages=c,c++,objc,f77

However, when the build (from make bootstrap-lean) attempts to configure
in libI77, we get

configuring in libI77
running /bin/sh /u/jbuck/gnu/src/cvs/tree-ssa/gcc/libf2c/libI77/configure  --host=i686-pc-linux-gnu --build=i686-pc-linux-gnu --enable-multilib --prefix --with-as=/usr/localdisk/binutils-2.13.1/bin/as --with-ld=/usr/localdisk/binutils-2.13.1/bin/ld --enable-languages=c,c++,objc,f77 --with-gcc-version-trigger=/u/jbuck/gnu/src/cvs/tree-ssa/gcc/gcc/version.c --with-libbanshee --with-target-subdir=i686-pc-linux-gnu --cache-file=../../config.cache --srcdir=/u/jbuck/gnu/src/cvs/tree-ssa/gcc/libf2c/libI77
configure: error: expected an absolute directory name for --prefix: --with-as=/usr/localdisk/binutils-2.13.1/bin/as

Notice that the value of the prefix option has been dropped, so the
--with-as switch is assumed to be the value of the prefix option.
This problem does not happen on the trunk or on the 3.3 branch (for
the identical configure line).

Re: tree-ssa branch: Fortran libraries don't build

[Diego Novillo]

On Mon, 2003-05-05 at 17:12, Joe Buck wrote:
> When trying to bootstrap the tree-ssa branch on a Red Hat 7.2 i686 system,
> the build of libI77 fails in the configure step.  The top-level configure
> line is
> 
> /u/jbuck/gnu/src/cvs/tree-ssa/gcc/configure \
>     --prefix /usr/localdisk/gcc-cvs/tree-ssa \
>     --with-as=/usr/localdisk/binutils-2.13.1/bin/as \
>     --with-ld=/usr/localdisk/binutils-2.13.1/bin/ld \
>     --enable-languages=c,c++,objc,f77
> 
Joe,

I just tried this after the latest merge and I can't reproduce the
failure.  We don't have any changes in libI77, so this may have been
fixed by the merge.  Could you try it again after updating your local
tree?


Thanks.  Diego.

Re: tree-ssa branch: Fortran libraries don't build

[Joe Buck]

On Wed, May 07, 2003 at 01:46:56PM -0400, Diego Novillo wrote:
> I just tried this after the latest merge and I can't reproduce the
> failure.  We don't have any changes in libI77, so this may have been
> fixed by the merge.  Could you try it again after updating your local
> tree?

I have a successful build now (the bug was probably in the makefile one
level up, but I'm not sure).

Test results are at 

http://gcc.gnu.org/ml/gcc-testresults/2003-05/msg00454.html

for C, C++, ObjC and Fortran.

There are more failures than for the trunk, but not vastly more.

By the way, the version isn't updated every day as it is on other branches
(so the title of the gcc-testresults mail says 20030504, not 0507).
Should this be fixed?  It makes the web archive of the list harder to
read (as all test results get put in the same thread).

Re: tree-ssa branch: Fortran libraries don't build

[Diego Novillo]

On Wed, 2003-05-07 at 20:57, Joe Buck wrote:

> I have a successful build now (the bug was probably in the makefile one
> level up, but I'm not sure).
> 
Good.  Thanks.

> By the way, the version isn't updated every day as it is on other branches
> (so the title of the gcc-testresults mail says 20030504, not 0507).
> Should this be fixed?  It makes the web archive of the list harder to
> read (as all test results get put in the same thread).
>
I manually update the version every time I merge with mainline.  Is
there any script that will the daily update automatically?  Or do people
just commit new changes to version.c every day?


Diego.

Re: tree-ssa branch: Fortran libraries don't build

[Andrew Pinski]

Look at maintainer-scripts/update_version.
I think the only thing you need to do is to add the tree-ssa branch 
name to `BRANCHES'.


Thanks,
Andrew Pinski

On Thursday, May 8, 2003, at 09:23 US/Eastern, Diego Novillo wrote:

> On Wed, 2003-05-07 at 20:57, Joe Buck wrote:
>
>> I have a successful build now (the bug was probably in the makefile 
>> one
>> level up, but I'm not sure).
>>
> Good.  Thanks.
>
>> By the way, the version isn't updated every day as it is on other 
>> branches
>> (so the title of the gcc-testresults mail says 20030504, not 0507).
>> Should this be fixed?  It makes the web archive of the list harder to
>> read (as all test results get put in the same thread).
>>
> I manually update the version every time I merge with mainline.  Is
> there any script that will the daily update automatically?  Or do 
> people
> just commit new changes to version.c every day?
>
>
> Diego.
>
>
>

Re: tree-ssa branch: Fortran libraries don't build

[Andreas Jaeger]

Diego Novillo <dnovillo@redhat.com> writes:

> I manually update the version every time I merge with mainline.  Is
> there any script that will the daily update automatically?  Or do people
> just commit new changes to version.c every day?

The script is:
gcc/maintainer-scripts/update_version

and it's installed on gcc.gnu.org (ask Gerald Pfeifer for details),

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

Re: tree-ssa branch: Fortran libraries don't build

[Gerald Pfeifer]

On Thu, 8 May 2003, Diego Novillo wrote:
> I manually update the version every time I merge with mainline.  Is
> there any script that will the daily update automatically?  Or do people
> just commit new changes to version.c every day?

This is fully automatic every since
  http://gcc.gnu.org/ml/gcc/1998-04/msg00952.html
which probably was my first serious use of sed. ;-)

If you update maintainer-scripts/update_version as others have suggested,
I'll happily install the updated version on gcc.gnu.org then.

Gerald
-- 
Gerald "Jerry"   pfeifer@dbai.tuwien.ac.at   http://www.pfeifer.com/gerald/

Re: tree-ssa branch: Fortran libraries don't build

[Diego Novillo]

On Thu, 2003-05-08 at 09:40, Gerald Pfeifer wrote:

> If you update maintainer-scripts/update_version as others have suggested,
> I'll happily install the updated version on gcc.gnu.org then.
> 
Excellent.  Thanks.  The change would have to be in mainline, I
suppose.  Is this patch OK?


Diego.


        * update_version (BRANCHES): Add tree-ssa branch.

Index: update_version
===================================================================
RCS file: /cvs/gcc/gcc/maintainer-scripts/update_version,v
retrieving revision 1.11
diff -d -u -p -r1.11 update_version
--- update_version      4 May 2003 22:51:26 -0000       1.11
+++ update_version      8 May 2003 13:56:50 -0000
@@ -17,6 +17,7 @@ $CVS co gcc/ChangeLog
 BRANCHES=`$CVS status -v gcc/ChangeLog \
           | awk '{print $1;}' \
           | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
+          | egrep 'tree-ssa-[0-9]+-branch$' \
           | egrep -v $IGNORE_BRANCHES`
 # Always update the mainline.
 BRANCHES="${BRANCHES} HEAD"

Re: tree-ssa branch: Fortran libraries don't build

[Andrew Pinski]

I do not think this will work since the pervious egrep will get rid of 
the branch name before the new one is ran,
The additional egrep will get rid of all the previous (release) 
branches.

Thanks,
Andrew Pinski

On Thursday, May 8, 2003, at 09:57 US/Eastern, Diego Novillo wrote:

> On Thu, 2003-05-08 at 09:40, Gerald Pfeifer wrote:
>
>> If you update maintainer-scripts/update_version as others have 
>> suggested,
>> I'll happily install the updated version on gcc.gnu.org then.
>>
> Excellent.  Thanks.  The change would have to be in mainline, I
> suppose.  Is this patch OK?
>
>
> Diego.
>
>
>         * update_version (BRANCHES): Add tree-ssa branch.
>
> Index: update_version
> ===================================================================
> RCS file: /cvs/gcc/gcc/maintainer-scripts/update_version,v
> retrieving revision 1.11
> diff -d -u -p -r1.11 update_version
> --- update_version      4 May 2003 22:51:26 -0000       1.11
> +++ update_version      8 May 2003 13:56:50 -0000
> @@ -17,6 +17,7 @@ $CVS co gcc/ChangeLog
>  BRANCHES=`$CVS status -v gcc/ChangeLog \
>            | awk '{print $1;}' \
>            | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
> +          | egrep 'tree-ssa-[0-9]+-branch$' \
>            | egrep -v $IGNORE_BRANCHES`
>  # Always update the mainline.
>  BRANCHES="${BRANCHES} HEAD"
>
>
>

Re: tree-ssa branch: Fortran libraries don't build

[Diego Novillo]

On Thu, May 08, 2003 at 10:03:04AM -0400, Andrew Pinski wrote:
> I do not think this will work since the pervious egrep will get rid of 
> the branch name before the new one is ran,
> The additional egrep will get rid of all the previous (release) 
> branches.
> 
Oh, right.  Serves me right for trying to play cute.  This one
based on brute force and ignorance should work.


Diego.

	* update_version (BRANCHES): Add tree-ssa branch.

Index: update_version
===================================================================
RCS file: /cvs/gcc/gcc/maintainer-scripts/update_version,v
retrieving revision 1.11
diff -d -u -p -r1.11 update_version
--- update_version	4 May 2003 22:51:26 -0000	1.11
+++ update_version	8 May 2003 14:19:00 -0000
@@ -14,10 +14,11 @@ CVS=${CVS:-/usr/local/bin/cvs}
 
 # Compute the branches which we should update.
 $CVS co gcc/ChangeLog
-BRANCHES=`$CVS status -v gcc/ChangeLog \
-          | awk '{print $1;}' \
-          | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
-          | egrep -v $IGNORE_BRANCHES`
+BRANCHES="`$CVS status -v gcc/ChangeLog \
+           | awk '{print $1;}' \
+           | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
+           | egrep -v $IGNORE_BRANCHES` \
+	   tree-ssa-20020619-branch"
 # Always update the mainline.
 BRANCHES="${BRANCHES} HEAD"
 

Re: tree-ssa branch: Fortran libraries don't build

[Alexandre Oliva]

Try this instead:

-          | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
+          | egrep '(gcc|tree-ssa)-[0-9_]+-branch$' \

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer

Re: tree-ssa branch: Fortran libraries don't build

[Joseph S. Myers]

On Thu, 8 May 2003, Alexandre Oliva wrote:

> Try this instead:
> 
> -          | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
> +          | egrep '(gcc|tree-ssa)-[0-9_]+-branch$' \

If extra active branches are to get automatic date dumps, I'd rather
having an explicit list of branches that get added rather than
complicating the egrep.  Or, better, move to an explicit list of active
branches to bump (as recently discussed in the context of not keeping
bumping dates on closed branches), but then creating a script to do some
of the work of making a new release branch would be better than adding yet
another step to the branching process.

-- 
Joseph S. Myers
jsm28@cam.ac.uk

Re: tree-ssa branch: Fortran libraries don't build

[Gerald Pfeifer]

On Thu, 8 May 2003, Joseph S. Myers wrote:
> If extra active branches are to get automatic date dumps, I'd rather
> having an explicit list of branches that get added rather than
> complicating the egrep.

I agree, and in fact that's what I'm planning to do.

Diego, I'll have to leave now, but I promise I'll take care of updating
and installing your patch on gcc.gnu.org tomorrow.

Gerald

Re: tree-ssa branch: Fortran libraries don't build

[Gabriel Dos Reis]

"Joseph S. Myers" <jsm28@cam.ac.uk> writes:

| On Thu, 8 May 2003, Alexandre Oliva wrote:
| 
| > Try this instead:
| > 
| > -          | egrep 'gcc-[0-9]+_[0-9]+-branch$' \
| > +          | egrep '(gcc|tree-ssa)-[0-9_]+-branch$' \
| 
| If extra active branches are to get automatic date dumps, I'd rather
| having an explicit list of branches that get added rather than
| complicating the egrep.  Or, better, move to an explicit list of active
| branches to bump (as recently discussed in the context of not keeping
| bumping dates on closed branches),

I don't recall if it were you or not but someone pointed out that it
was a feature that active branches got automatic date bump without
explicitly giving a list of such branches, hence the "grep -v".  
Now, you're contemplating the other way.  I'm confused.

|  but then creating a script to do some
| of the work of making a new release branch would be better than adding yet
| another step to the branching process.

I'm for anything that can simplify life.

-- Gaby

Re: tree-ssa branch: Fortran libraries don't build

[Diego Novillo]

On Thu, 2003-05-08 at 11:29, Gerald Pfeifer wrote:

> Diego, I'll have to leave now, but I promise I'll take care of updating
> and installing your patch on gcc.gnu.org tomorrow.
> 
Sure.  I'm in no hurry.


Thanks.  Diego.

Re: tree-ssa branch: Fortran libraries don't build

[Joseph S. Myers]

On Thu, 8 May 2003, Gabriel Dos Reis wrote:

> I don't recall if it were you or not but someone pointed out that it
> was a feature that active branches got automatic date bump without
> explicitly giving a list of such branches, hence the "grep -v".  
> Now, you're contemplating the other way.  I'm confused.

It was designed as a feature - to reduce the number of steps involved in
making a new release branch, not for an underlying philosophical advantage
of one way over another.  I'm saying that if we change the method to an
explicit list - including other non-release branches in that same list
rather than a grep then adding them on - then this should be done in
conjunction with a branching script so that again the number of steps to
create a release branch is reduced (i.e., the branching script should
update the list of active branches, and several other places that at
present need updating manually as detailed in branching.html).

-- 
Joseph S. Myers
jsm28@cam.ac.uk