Thread: [Module-build-general] Ready for 0.19?
Status: Beta
Brought to you by:
kwilliams
Menu
▾
▴
module-build-general
|
From: Ken W. <ke...@ma...> - 2003-07-03 20:55:04
|
Hi, I've released a new beta, 0.18_04, to cpan and sourceforge. It gets the 'install_base' parameter working, completing the design document I posted here a few days ago. I'd love it if people could test it out, by either/both of running the regression tests & using it to test/install their own Module::Build-using modules. If it looks good, I'll make the 0.19 release. -Ken |
|
From: Dave R. <au...@ur...> - 2003-07-04 04:05:26
|
On Thu, 3 Jul 2003, Ken Williams wrote:
> I've released a new beta, 0.18_04, to cpan and sourceforge. It gets
> the 'install_base' parameter working, completing the design document I
> posted here a few days ago. I'd love it if people could test it out,
> by either/both of running the regression tests & using it to
> test/install their own Module::Build-using modules.
Passes all tests on Debian unstable with 5.8.0 and 5.6.1, but there are
some small problems with 5.00503.
When I ran Build.PL for the distro, it told me I needed Module::Info.
So I went ahead and installed it, which worked fine under 5.00503.
However, after doing this I get these test results:
t/basic.........ok
t/compat........ok 16/23Argument "0.18_04" isn't numeric in entersub at (eval 4) line 3.
t/compat........ok
t/runthrough....ok 8/26B::Module::Info,packages use failed with 255 saying:
use of backend Module::Info failed: Perl 5.006 required--this is only version 5.00503, stopped at /usr/local/perl5.00503/lib/site_perl/5.005/B/Utils.pm line 3.
BEGIN failed--compilation aborted at /usr/local/perl5.00503/lib/site_perl/5.005/B/Utils.pm line 3.
BEGIN failed--compilation aborted at /usr/local/perl5.00503/lib/site_perl/5.005/B/Module/Info.pm line 7.
BEGIN failed--compilation aborted at (eval 1) line 2.
at /root/cpan-5.00503/build/Module-Build-0.18_04/t/Sample/lib/Sample.pm line 0
BEGIN failed--compilation aborted.
t/runthrough....FAILED tests 17-18, 21-22
Failed 4/26 tests, 84.62% okay
t/signature.....skipped
all skipped: Skipping unless $ENV{TEST_SIGNATURE} is true
t/xs............ok
Failed Test Stat Wstat Total Fail Failed List of Failed
-------------------------------------------------------------------------------
t/runthrough.t 26 4 15.38% 17-18 21-22
1 test skipped.
Failed 1/5 test scripts, 80.00% okay. 4/75 subtests failed, 94.67% okay.
The above errors are not the cause of the test failures, but they're
annoying and will not inspire end-user confidence ;)
Maybe just don't bother trying to load Module::Info "unless $] >= 5.006"?
Here's the verbose output for the failing tests:
not ok 17
# Test 17 got: 'Can't figure out where to install things of type 'script' at /root/cpan-5.00503/build/Module-Build-0.18_04/blib/lib/Module/Build/Base.pm line 1439.
' (t/runthrough.t at line 100)
# Expected: ''
Should have installed module as /root/cpan-5.00503/build/Module-Build-0.18_04/t/install_test/usr/local/perl5.00503/lib/site_perl/5.005/Sample.pm
not ok 18
# Failed test 18 in t/runthrough.t at line 105
Installing /root/cpan-5.00503/build/Module-Build-0.18_04/t/install_test/usr/local/perl5.00503/lib/5.00503/Sample.pm
Installing /root/cpan-5.00503/build/Module-Build-0.18_04/t/install_test/usr/local/perl5.00503/bin/script
ok 19
Should have installed module as /root/cpan-5.00503/build/Module-Build-0.18_04/t/install_test/usr/local/perl5.00503/lib/5.00503/Sample.pm
ok 20
not ok 21
# Test 21 got: 'Can't figure out where to install things of type 'script' at /root/cpan-5.00503/build/Module-Build-0.18_04/blib/lib/Module/Build/Base.pm line 1439.
' (t/runthrough.t at line 117)
# Expected: ''
Should have installed module as /root/cpan-5.00503/build/Module-Build-0.18_04/t/install_test/foo/bar/Sample.pm
not ok 22
# Failed test 22 in t/runthrough.t at line 120
The problem seems to be that my Config.pm does not define anything for
"installsitebin".
I have no idea why that should be the case. For versions 5.00503 and
5.6.1 of Perl on my machine, I compiled them myself, while 5.8.0 is
the Debian packaged version, so I suspect this is a general 5.00503
problem. I think you may need to just use "installbin" on 5.00503.
With the patch below my sig, all tests pass with 5.00503, though the
annoying noise from Module::Info persists.
-dave
/*=======================
House Absolute Consulting
www.houseabsolute.com
=======================*/
--- lib/Module/Build/Base.pm~ 2003-07-03 15:40:46.000000000 -0500
+++ lib/Module/Build/Base.pm 2003-07-03 23:04:26.000000000 -0500
@@ -99,7 +99,7 @@
lib => $c->{installsitelib},
arch => $c->{installsitearch},
bin => $c->{installsitebin},
- script => $c->{installsitescript} || $c->{installsitebin},
+ script => $c->{installsitescript} || $c->{installsitebin} || $c->{installbin},
bindoc => $c->{installsiteman1dir},
libdoc => $c->{installsiteman3dir},
},
@@ -107,7 +107,7 @@
lib => $c->{installvendorlib},
arch => $c->{installvendorarch},
bin => $c->{installvendorbin},
- script => $c->{installvendorscript} || $c->{installvendorbin},
+ script => $c->{installvendorscript} || $c->{installvendorbin} || $c->{installbin},
bindoc => $c->{installvendorman1dir},
libdoc => $c->{installvendorman3dir},
},
|
|
From: Ken W. <ke...@ma...> - 2003-07-07 02:04:17
|
On Thursday, July 3, 2003, at 11:04 PM, Dave Rolsky wrote: > On Thu, 3 Jul 2003, Ken Williams wrote: > >> I've released a new beta, 0.18_04, to cpan and sourceforge. It gets >> the 'install_base' parameter working, completing the design document I >> posted here a few days ago. I'd love it if people could test it out, >> by either/both of running the regression tests & using it to >> test/install their own Module::Build-using modules. > > Passes all tests on Debian unstable with 5.8.0 and 5.6.1, but there are > some small problems with 5.00503. Both ideas sound good to me, I'll apply them to CVS. -Ken |
|
From: Ken W. <ke...@ma...> - 2003-07-07 02:14:33
|
On Sunday, July 6, 2003, at 09:04 PM, Ken Williams wrote:
>
> On Thursday, July 3, 2003, at 11:04 PM, Dave Rolsky wrote:
>
>> On Thu, 3 Jul 2003, Ken Williams wrote:
>>
>>> I've released a new beta, 0.18_04, to cpan and sourceforge. It gets
>>> the 'install_base' parameter working, completing the design document
>>> I
>>> posted here a few days ago. I'd love it if people could test it out,
>>> by either/both of running the regression tests & using it to
>>> test/install their own Module::Build-using modules.
>>
>> Passes all tests on Debian unstable with 5.8.0 and 5.6.1, but there
>> are
>> some small problems with 5.00503.
>
> Both ideas sound good to me, I'll apply them to CVS.
Say, Dave: it seems like it would be better to default 'script' like so:
$c->{installsitescript} || $c->{installsitebin} ||
$c->{installscript},
rather than
$c->{installsitescript} || $c->{installsitebin} || $c->{installbin},
Do you have an 'installscript' on 5.005?
-Ken
|
|
From: Dave R. <au...@ur...> - 2003-07-07 02:18:26
|
On Sun, 6 Jul 2003, Ken Williams wrote:
> >> Passes all tests on Debian unstable with 5.8.0 and 5.6.1, but there
> >> are
> >> some small problems with 5.00503.
> >
> > Both ideas sound good to me, I'll apply them to CVS.
>
> Say, Dave: it seems like it would be better to default 'script' like so:
>
> $c->{installsitescript} || $c->{installsitebin} ||
> $c->{installscript},
>
> rather than
>
> $c->{installsitescript} || $c->{installsitebin} || $c->{installbin},
>
> Do you have an 'installscript' on 5.005?
Yep, I do. I just didn't notice it ;)
There's a lot of entries in that Config.pm file ;)
-dave
/*=======================
House Absolute Consulting
www.houseabsolute.com
=======================*/
|
|
From: Mathieu A. <ma...@ma...> - 2003-07-04 06:37:14
|
+-Le 03/07/2003 23:04 -0500, Dave Rolsky =E9crivait : | When I ran Build.PL for the distro, it told me I needed Module::Info. | So I went ahead and installed it, which worked fine under 5.00503. | However, after doing this I get these test results: I believe there are a few remnants in Module::Info, there is a use 5.006 and an "our" just after. I guess that removing the "our"s, adding a use vars at the beginning, and also adding a dependency on File::Spec 0.82 will be good, because the one with perl 5.00503 does not have rel2abs for instance. --=20 Mathieu Arnold |
|
From: Dave R. <au...@ur...> - 2003-07-04 06:40:00
|
On Fri, 4 Jul 2003, Mathieu Arnold wrote: > > > +-Le 03/07/2003 23:04 -0500, Dave Rolsky =E9crivait : > | When I ran Build.PL for the distro, it told me I needed Module::Info. > | So I went ahead and installed it, which worked fine under 5.00503. > | However, after doing this I get these test results: > > I believe there are a few remnants in Module::Info, there is a use 5.006 > and an "our" just after. I guess that removing the "our"s, adding a use > vars at the beginning, and also adding a dependency on File::Spec 0.82 wi= ll > be good, because the one with perl 5.00503 does not have rel2abs for > instance. No, the big problem is that most of the functionality requires the B::* modules. I think the fix is to simply not load it with 5.00503. -dave /*=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D House Absolute Consulting www.houseabsolute.com =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D*/ |
|
From: <ajs...@ya...> - 2003-07-04 09:24:19
|
Ken Williams <ke...@ma...> wrote:
> I've released a new beta, 0.18_04, to cpan and sourceforge. It gets
> the 'install_base' parameter working, completing the design document
> I posted here a few days ago. I'd love it if people could test it
> out, by either/both of running the regression tests & using it to
> test/install their own Module::Build-using modules.
>
> If it looks good, I'll make the 0.19 release.
Still passes OK on Windows.
Just a random thought. I noticed t/compat.t contains:
skip_test("Don't know how to invoke 'make'") unless $Config{make};
which is good. But what if $Config{make} points to something that
is not there? (I fear some Windows users may have $Config{make}
set to 'nmake' yet not have nmake installed).
Given that Module::Build does not depend on make, I think it would
be nice if compat.t could check if $Config{make} was indeed
available on the PATH and skip the tests if not -- instead of
failing, as it currently does. Does anyone know of a clean,
portable way to tell if a command is available on the PATH?
If not, a simple workaround is to start compat.t with the test:
$Config{make} force_do_it
Remember, force_do_it is a dummy target that does nothing, so this
command should always succeed with a return code of zero ... unless
$Config{make} is indeed unavailable.
/-\
http://mobile.yahoo.com.au - Yahoo! Mobile
- Check & compose your email via SMS on your Telstra or Vodafone mobile.
|
Thanks for helping keep SourceForge clean.
X