Developers of Sympa

[Olivier Salaün <address@concealed>]

Hello Guillaume,

Sorry we did not take your work into account earlier, just because lack 
of time....

We are aware that the way we are using the autotools is far from perfect 
and your involvment in this task is a good occasion for us to make the 
Sympa installation scripts a bit cleaner.
Apart from the satisfaction of using the autotools the proper way, here 
are the goals we'd like to achieve while performing the changes :

  1. make the scripts more maintainable. Currently all Makefile include
     lots of duplicate variable definitions that we'd like to get rid
     of, ie define them once only.
  2. make it easier for contributors/packagers to understand Sympa
     installation process. This is the main advantage of doing things
     the "straight" way
  3. keep things working. Over time, the installation scripts have
     become a spaghetti plate, however it does the installation job and
     we need to validate any changes before committing changes !

Step (3) should be easier to achieve, given your latest set of atomic 
patches, whereas the first 3 ones mix lots of changes and are therefore 
more difficult to check on our end.
However, let me comment all your change proposals below :

Guillaume Rousse a écrit :
> Guillaume Rousse a écrit :
>   
> > Hello developpers.
> >
> > I'm mandriva sympa package maintainer. We met during JRES, and I told
> > you how awfull sympa autotools usage was...
> >
> > Here are two first patches attempting to change the situation, and move
> > toward a standard use of those tools.
> >
> > The first one (autoconf-cleanup) doesn't change much in the installation
> > process, it's rather a purely syntaxic cleanup:
> > - rename configure.in to configure.ac
> >     
ok
> > - define software version inside configure.ac directly
> >     
Currently we are maintaining in a separate .version file.
This file is later used to

  1. include the current version in Version.pm
  2. compare installed version and to-be-installed versions during an
     upgrade and print significant changes during the "make install"
     (via the important_changes.pl script)

It was convenient to simply update this .version file while preparing a 
new Sympa release, however we might change our habits if needed.

Could you please explain why putting the Sympa version number in 
configure.ac is a required change ?
> > - use current AC_INIT and AC_OUTPUT syntax
> > - define autoconf 2.59 as minimal version
> > - wrap long lines
> > - use AC_HELP_STRING for formatting options help output
> > - make all binaries tests use the same logic, meaning no default value,
> > and only look in PATH if no value given by user
> > - don't call AC_SUBST for variables set through AC_PATH_PROG (done
> > implicitely)
> >     
ok
> > - don't put explicit substitutions targets in Makefile.am, automake does
> > it automatically
> >     
Can you explain?
> > - removal of AM_MAINTAINER_MODE in configure.ac (it prevent automatic
> > makefile rebuild/configure re-invocation)
> >     
ok
> > The second one (use-automake-correctly), bring real change. It is not a
> >  finished jobs (too many issues to discuss), but it shows how automake
> > is supposed to be used. So:
> > - removal of all targets already defined by automake (install, all, etc..)
> > - removal of all recursion targets (makedoc, makeman,...)
> >     
This seems to be a legitimate change, however note that it requires a 
way to ensure we don't loose any targets, ie lots of testing.
> > - conversion of all static Makefile in src and doc trees to Makefile.ams
> > - definition of SUBDIR variable in top-level, src and doc subdirs
> > Makefile.am
> > - definition of primary targets in those makefile, as foo_BAR variables,
> > where foo is the installation directory, and BAR the type of the file to
> > install
> >     
ok
> > - definition of EXTRA_DIST and CLEANFILES variable to ensure make clean
> > and make dist will work as expected
> >     
Can you please explain which files will be declared as EXTRA_DIST?
> > - use a standard in-place substitution system for variables defined by
> > configure process, as per
> > http://www.gnu.org/software/autoconf/manual/autoconf-2.57/html_mono/autoconf.html#SEC24
> >     
The current use of the subst.pl script is pretty satisfactory ; it's 
flexible and the script also allows us to do additional changes like 
commenting do_log() calls that can be skipped (performances issue). The 
only drawback of this substitution process (appart its non standard 
character) is that we need to set lots of (error-prone) environment 
variables in Makefiles before each call.
The new substitution system that you refer to seems less flexible (you 
need to define a twisted sed expression in the Makefile and extend this 
engine for each new variable) and seems to oblige us to add the '.in' 
extension to perl modules, which is not convenient to us.
> > - drop all reference to custom variables in configure.ac for those
> > directories defined in GNU's standards
> >     
Can you precise which are these custom variables?
> > - drop non-standard /home/sympa prefix
> >     
This prefix has been used for 10 years now, therefore it's not so simple 
to change it because it would break existing installations during upgrades.
What is the reason for having a different default ? Which one do you 
recommend ?
> > - usage of a config.h headers rather than command line flags to pass C
> > symbols for building binaries
> >     
Please explain why this change is needed?
> > I didn't finished the work mainly because setup is unclear. Original
> > configure script defines a lots of different subdirectories with similar
> > content type, whereas most software usually just define directories for:
> > - configuration ($prefix/etc/sympa by default)
> > - non-variable content ($prefix/share/sympa by default)
> > - variable content ($prefix/var/lib/sympa by default)
> > - documentation ($prefix/share/doc/sympa by default)
> > and then install content relatively to those top-level directories.
> >     
We had this change in mind, ie gathering all perl modules in a lib/ 
directory, all daemons in sbin/, etc.
However I'd consider this a second time work, after all the other 
cleaning and simplification has been done.
> > I didn't configured installation, nor ensured substitutions were
> > correctly handled for those destination directories not cleary matching
> > this standard setup (samples and user config).
> >
> > Also, you'll notice how performing a lot of substitutions in perl
> > modules is cumbersome. You'd rather have a unique .pm files defining
> > those constants (using constant pragma, or ReadOnly module), sourced by
> > all others, so as to substitude only once.
> >     
It could be a good idea to get rid of all --VAR-- references in the 
code, however these are precisely data that are part of the bootstrap 
(--PERL-- refers to the path to the perl interpreter, --LIBDIR-- tells 
where to find other required perl modules, --CONFIG-- tells where the 
main configuration file stands, etc). Therefore we need these 
"bootstrap" information to be provided quite early in the process.
> > And I didn't handled the po subdirectory neither. Not very difficult to
> > do, but the procedure to do it is a bit different from other contents,
> > so I wanted first to demonstrate pure automake usage. For more
> > explanations, I suggest reading http://www.lrde.epita.fr/~adl/autotools.html
> >
> >     
Note that the po/Makefile has been changed by David in the meanwhile, 
just in case you're working on it.
> > Hope you'll appreciate the work  :) 
> >     
Of course we do ; you're providing time and expertise that we don't have 
. Moreover it's useful to have an external point of view of somebody who 
asks why when things are too twisted :)
> I realized just after sending this mail the patch was broken, as it
> didn't included removed and renamed files.
>
> Here is a fixed subset of previous patch, dealing with doc directory
> only. The list of changes follows:
> - conversion of static Makefile in doc and doc/man8 to
> automake-generated makefiles
>   
ok
> - removal of installation targets (installdoc, installman) in top-level
> Makefile, and usage of primary variable (doc_DATA, man_MANS) to let
> automake handle them
>   
I could not find any definition of doc_DATA in the provided patches...
> - removal of recursion targets (documentation, man) in top-level
> Makefile, and usage of SUBDIR variable to let automake handle it
>   
> - removal of all target in top-level Makefile, to let automake handle it
> - removal of clean target in top-level Makefile, and usage of CLEANFILES
> variable to met automake handle it
> - definition of EXTRA_DEST variable to correctly include file in archive
> generated by 'make dist'
> - removal of .latex2html-init file, as it is not handled by build process
>   
ok
> - removal of man pages, as they are generated from pod files
>   
> - test for pod2man in configure, and fatal error if not found
>
>   
ok
> I hope this is easier to understand.
>   

Next week I'll manage to test the last set of patches you provided in 
your last email and I'll let you know about the results.
If things work fine, I'll commit the changes in our SVN.

Thanks again for your help Guillaume.