Perl code coverage

[Ralf Wildenhues]

I recently discovered Devel::Cover (Debian package libdevel-cover-perl).
This rocks.  Kudos to Paul Johnson for writing this module.

To use Devel::Cover, I have started a branch perl-coverage (off of
maint) which introduces a few helper targets, e.g.,
  make check-coverage

which is veeery slow, but also quite interesting.

I will follow up on automake-patches with the patches in the branch, as
far as I've merged it into master.  (Please speak up if you'd like me to
push the branch to savannah as well; otherwise, you can easily get at it
with something like merge-point^2.)

Up to now, I've tackled the trivial pod coverage, some Perl module
coverage, and started on 'automake' script coverage, beginning at the
end of the script.  A big part of it is fairly trivial, but IMHO still
worthwhile to add since it allows us to be more confident about
correctness when changing the code later.  I've found a few likely bugs
with this already, too, and the peaks in the 'time' column below whisper
'missed-optimization'.


Use of Devel::Cover causes a few test failures; notably, it doesn't work
with ithreads, and it causes some output even with -silent.  It might be
necessary to not close STDOUT in General::END, not sure yet about that.

In summary, running the Automake testsuite on a GNU/Linux x86 system
shows roughly the following coverage (I've had a couple of minor,
unfinished patches in the tree).  For a description of the data, see:
<http://search.cpan.org/dist/Devel-Cover/lib/Devel/Cover/Tutorial.pod>.
Devel::Cover also exports very nice per-file, per-statement etc. HTML
files, I'm not sure whether it is worthwhile to put them up somewhere.

----------------------------------- ------ ------ ------ ------ ------ ------ ------
File                                  stmt   bran   cond    sub    pod   time  total
----------------------------------- ------ ------ ------ ------ ------ ------ ------
...make/lib/Automake/ChannelDefs.pm   93.0   92.9    n/a   88.2   88.9    0.2   92.2
...utomake/lib/Automake/Channels.pm   83.4   66.1   71.4   86.8  100.0    1.0   79.5
...tomake/lib/Automake/Condition.pm   97.8   94.8  100.0   96.2   94.7    2.2   97.1
...ake/lib/Automake/Configure_ac.pm   96.4   87.5   50.0  100.0    0.0    0.0   93.3
...e/lib/Automake/DisjConditions.pm   78.7   70.0    n/a   84.2  100.0    1.6   77.6
...tomake/lib/Automake/FileUtils.pm   56.9   31.2   18.2   76.0  100.0    0.2   50.0
...automake/lib/Automake/General.pm  100.0  100.0    n/a  100.0    0.0    0.2  100.0
...ke/automake/lib/Automake/Item.pm  100.0   50.0   66.7  100.0  100.0    1.7   94.1
...automake/lib/Automake/ItemDef.pm  100.0    n/a    n/a  100.0  100.0    0.4  100.0
...utomake/lib/Automake/Location.pm   63.5    n/a   25.0   80.0    0.0    0.9   63.6
...automake/lib/Automake/Options.pm   94.8   92.5   92.3   96.0  100.0    0.1   93.7
...ke/automake/lib/Automake/Rule.pm   93.6   90.0   81.2   89.2  100.0    1.5   91.1
...automake/lib/Automake/RuleDef.pm  100.0    n/a    n/a  100.0   33.3    0.1  100.0
.../automake/lib/Automake/Struct.pm   79.0   51.9   44.4   82.9   50.0    0.4   67.7
.../automake/lib/Automake/VarDef.pm   97.6   85.7  100.0  100.0   90.0    0.6   96.7
...utomake/lib/Automake/Variable.pm   95.7   86.8   83.8   96.4  100.0    3.0   92.3
...automake/lib/Automake/Version.pm   96.6   75.0   83.3  100.0  100.0    0.0   87.9
...ke/automake/lib/Automake/Wrap.pm  100.0  100.0  100.0  100.0   66.7    0.4  100.0
...e/automake/lib/Automake/XFile.pm   71.0   37.5   14.3   77.8   55.6   61.2   59.2
.../lib/Automake/tests/Condition.pl   87.9   50.0   33.3  100.0    n/a    0.0   75.5
...Automake/tests/DisjConditions.pl   82.3   50.0   44.4  100.0    n/a    0.0   72.1
...ke/lib/Automake/tests/Version.pl   66.7   50.0    n/a  100.0    n/a    0.0   65.6
...omake/lib/Automake/tests/Wrap.pl   85.7   50.0    n/a  100.0    n/a    0.0   83.3
...nload/cvs/automake/build/aclocal   97.9   93.7   72.3   97.5    n/a   17.9   94.4
...load/cvs/automake/build/automake   96.1   89.2   87.3   96.7    n/a    6.3   93.2
...ake/build/lib/Automake/Config.pm  100.0    n/a    n/a  100.0    n/a    0.0  100.0
Total                                 92.8   83.0   81.6   93.3   86.6  100.0   89.2
----------------------------------- ------ ------ ------ ------ ------ ------ ------

[Stefan Bienert-2]

Hi list!

Is there a way to stop automake producing a rule for dvi docs?
It just makes my 'distcheck' fail.

kind regards,

Stefan

[Ralf Wildenhues]

* Stefan Bienert wrote on Mon, Oct 19, 2009 at 12:45:19AM CEST:
> Is there a way to stop automake producing a rule for dvi docs?
> It just makes my 'distcheck' fail.

dvi:

Cheers,
Ralf

[Stefan Bienert-2]

Ralf Wildenhues wrote:
> * Stefan Bienert wrote on Mon, Oct 19, 2009 at 12:45:19AM CEST:
>> Is there a way to stop automake producing a rule for dvi docs?
>> It just makes my 'distcheck' fail.
>
> dvi:

This does not work:

cd ../.. && /bin/sh
/home/user/projects/documentation.git/share/misc/config/missing --run
automake-1.10 --gnu  share/doc/Makefile
share/doc/Makefile.am:76: user target `dvi' defined here...
automake-1.10: ... overrides Automake target `dvi' defined here
share/doc/Makefile.am:76: consider using dvi-local instead of dvi

I also tried dvi-local, but this does not work either!

kind regards,

Stefan

--
Stefan Bienert
Zentrum fuer Bioinformatik
Universitaet Hamburg
Bundesstrasse 43
20146 Hamburg
Germany

Email: bienert@...
Phone:  [phone]
Fax:    [phone]

[Ralf Wildenhues]

* Stefan Bienert wrote on Mon, Oct 19, 2009 at 02:06:40PM CEST:
> cd ../.. && /bin/sh
> /home/user/projects/documentation.git/share/misc/config/missing --run
> automake-1.10 --gnu  share/doc/Makefile
> share/doc/Makefile.am:76: user target `dvi' defined here...
> automake-1.10: ... overrides Automake target `dvi' defined here
> share/doc/Makefile.am:76: consider using dvi-local instead of dvi

This is not an error, merely a warning.  You can silence it with
  AUTOMAKE_OPTIONS = -Wno-override

> I also tried dvi-local, but this does not work either!

That's expected: dvi-local exists to amend the dvi rule.

Cheers,
Ralf

[Stefan Bienert-2]

Splendid!

Thanks mate, this gives me exactly the behaviour I was seeking, while
exactly means E-X-A-C-T-L-Y.

I didn't favour a global change of rules but a local one...mission
accomplished!

I just feel a little bit dumb, because I searched the "Changing
Automakes behaviour" section by myself, before...and didn't get it.

kind regards,

Stefan

Ralf Wildenhues wrote:

> * Stefan Bienert wrote on Mon, Oct 19, 2009 at 02:06:40PM CEST:
>> cd ../.. && /bin/sh
>> /home/user/projects/documentation.git/share/misc/config/missing --run
>> automake-1.10 --gnu  share/doc/Makefile
>> share/doc/Makefile.am:76: user target `dvi' defined here...
>> automake-1.10: ... overrides Automake target `dvi' defined here
>> share/doc/Makefile.am:76: consider using dvi-local instead of dvi
>
> This is not an error, merely a warning.  You can silence it with
>   AUTOMAKE_OPTIONS = -Wno-override
>
>> I also tried dvi-local, but this does not work either!
>
> That's expected: dvi-local exists to amend the dvi rule.
>
> Cheers,
> Ralf
>