[BUG] configure problem on the arm architecture

[gerd]

With the Fedora Build System I tried to build Parrot on the ARM architecture ( http://en.wikipedia.org/wiki/ARM_architecture).

The configuration stops wih a byteorder problem.

URL:  http://arm.koji.fedoraproject.org/koji/getfile?taskID=96590&name=build.log

auto::byteorder - Compute native byteorder for wordsize... step auto::byteorder died during execution: Unsupported byte-order [126291410274]! at config/auto/byteorder.pm line 69.

at Configure.pl line 75

auto::va_ptr - Test the type of va_ptr...........................stack.

The second time I modified the file config/auto/byteorder.pm to use 'little-endian' like that:

sub _evaluate_byteorder {

my ($self, $conf, $byteorder) = @_; #if ( $byteorder =~ /^{1234/ ) {}

$conf->data->set(

byteorder => $byteorder, bigendian => 0

); $self->set_result('little-endian');

return 1;

}

With that modification the next build ( http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=96592) works.

How to patch config/auto/byteorder.pm that it works with the arm architecture also?

[gerd]

The patch added at  http://trac.parrot.org/parrot/changeset/43468 fix it.

[doughera]

Replying to gerd:

The patch added at  http://trac.parrot.org/parrot/changeset/43468 fix it.

That patch nicely works around the problem, but it would still be nice to know why the failure happened in the first place and fix the test more generally so that such workarounds wouldn't be necessary.

Could the original poster run that Configure.pl test in verbose mode and help figure out why it failed?

[gerd]

I produced the output of "perl Configure.pl --test --verbose". The URL to see it is:  http://arm.koji.fedoraproject.org/koji/getfile?taskID=98661&name=build.log Please ignore the first Configure.pl output in this page.

[doughera]

Replying to gerd:

I produced the output of "perl Configure.pl --test --verbose". The URL to see it is:  http://arm.koji.fedoraproject.org/koji/getfile?taskID=98661&name=build.log Please ignore the first Configure.pl output in this page.

I don't understand this at all. The first Configure.pl output on this page shows it working fine. Was this perhaps patched with the change in r43468? You'd have to back out that patch for the test to be meaningful. Or is that why I'm supposed to ignore it?

The second report was generated with a bare

perl Configure.pl --test --verbose

which doesn't include any of the arm-specific flags that likely caused the problem in the first place, so I don't know what to make of that one either.

[gerd]

Hi, the previous Configure was with the work around. Now I took revision 43584 removed the work around and produced the output of "perl Configure.pl --test --verbose". The URL to see it is:  http://arm.koji.fedoraproject.org/koji/getfile?taskID=100840&name=build.log

[jkeenan]

Replying to gerd:

 http://arm.koji.fedoraproject.org/koji/getfile?taskID=100840&name=build.log

Following up on this ticket tonight, I tried to follow that URL. I got:

An error has occurred in the web interface code. This could be due 
to a bug or a configuration issue.
GenericError: no file "build.log" output by task 100840 

[jkeenan]

Replying to gerd:

Hi,

Gerd, as previously reported, I get an error when I go to this report:

> http://arm.koji.fedoraproject.org/koji/getfile?taskID=100840&name=build.log

Can you update us on the status of this problem?

Thank you very much.

kid51

[gerd]

For me the ticket can be closed. Since the patch is applied I do not hear any problems to build parrot on the arm architecture. Version 2.2.0 is the last build of Parrot on arm and Fedora.

 http://arm.koji.fedoraproject.org/koji/packageinfo?packageID=5461

[jkeenan]

Replying to gerd:

For me the ticket can be closed.

Closing. Thanks for looking into this.