From @cpansprout

I found this change to bisect-runner.pl very helpful. Is applying this a good idea? I will go ahead in a few days if nobody objects.

--

Father Chrysostomos

From @cpansprout

From​: Father Chrysostomos <sprout@​cpan.org>

bisect-runner.pl​: Don’t run tests for prereq modules

With the --module option, we *are* trying to see whether the module’s
tests pass. With the --with-module option, the modules are installed
merely as prerequisites for whatever program we are trying to test.
So skip the tests for the latter option, and make bisects faster.

diff --git a/Porting/bisect-runner.pl b/Porting/bisect-runner.pl
index 360c186..531891e 100755
--- a/Porting/bisect-runner.pl
+++ b/Porting/bisect-runner.pl
@@ -590,7 +590,8 @@ And then:
 --with-module module1,module2,...
 
 Like I<--module> above, except this simpl/private/tmp/open_8kdnmd43.txty installs the requested
-modules and they can then be used in other tests.
+modules and they can then be used in other tests.  The modules are
+installed without their tests being run, for speed.
 
 For example:
 
@@ -1508,7 +1509,14 @@ if ($options{module} || $options{'with-module'}) {
   for (@m) {
     s/-/::/g if /-/ and !m|/|;
   }
-  my $install = join ",", map { "'$_'" } @m;
+  my $install;
+  if ($options{module}) {
+    $install = join ",", map { "'$_'" } @m;
+    $install = "install($install)";
+  } else {
+    $install = join ",", map { "'$_'" } "install", @m;
+    $install = "notest($install)";
+  }
   my $last = $m[-1];
   my $shellcmd = "install($install); die unless CPAN::Shell->expand(Module => '$last')->uptodate;";
 

From @tonycoz

On Mon May 30 13​:30​:45 2016, sprout wrote​:

I found this change to bisect-runner.pl very helpful. Is applying
this a good idea? I will go ahead in a few days if nobody objects.

It really should test and skip if the test fails.

Tony

The RT System itself - Status changed from 'new' to 'open'

From @cpansprout

On Mon May 30 16​:37​:38 2016, tonyc wrote​:

On Mon May 30 13​:30​:45 2016, sprout wrote​:

I found this change to bisect-runner.pl very helpful. Is applying
this a good idea? I will go ahead in a few days if nobody objects.

It really should test and skip if the test fails.

First, I don’t know bisect-runner.pl or git-bisect well enough to make that change; secondly, it would not have helped for my use case, since I was getting module failures adjacent to the commit responsible for fixing the crash I’m trying to track.

So, perhaps for my use case I could add a separate --no-module-tests option. Any objections to that?

--

Father Chrysostomos

From @tonycoz

On Mon, May 30, 2016 at 05​:50​:01PM -0700, Father Chrysostomos via RT wrote​:

On Mon May 30 16​:37​:38 2016, tonyc wrote​:

On Mon May 30 13​:30​:45 2016, sprout wrote​:

I found this change to bisect-runner.pl very helpful. Is applying
this a good idea? I will go ahead in a few days if nobody objects.

It really should test and skip if the test fails.

First, I don’t know bisect-runner.pl or git-bisect well enough to make that change; secondly, it would not have helped for my use case, since I was getting module failures adjacent to the commit responsible for fixing the crash I’m trying to track.

So, perhaps for my use case I could add a separate --no-module-tests option. Any objections to that?

No, that sounds fine to me.

Tony

From @cpansprout

On Mon May 30 19​:01​:20 2016, tonyc wrote​:

On Mon, May 30, 2016 at 05​:50​:01PM -0700, Father Chrysostomos via RT
wrote​:

So, perhaps for my use case I could add a separate --no-module-tests
option. Any objections to that?

No, that sounds fine to me.

I have added such an option in commit 6830146.

--

Father Chrysostomos

@cpansprout - Status changed from 'open' to 'resolved'

From @wolfsage

On Tue, May 31, 2016 at 9​:36 AM, Father Chrysostomos via RT
<perlbug-followup@​perl.org> wrote​:

On Mon May 30 19​:01​:20 2016, tonyc wrote​:

On Mon, May 30, 2016 at 05​:50​:01PM -0700, Father Chrysostomos via RT
wrote​:

So, perhaps for my use case I could add a separate --no-module-tests
option. Any objections to that?

No, that sounds fine to me.

I have added such an option in commit 6830146.

Yes that sounds better to me.

Thanks!

-- Matthew Horsfall (alh)