问题
I am trying to install a conda package of a Perl module. So far I'm able to create the package using conda-build. For that I have a recipe containing a build.sh and a meta.yaml files.
I then install it using conda-install in a new environment, I'd like to be able able to run some Perl scripts located in the Perl module I just installed.
All those steps work well but when I'm running some scripts I have an error saying :
Can't locate PMP/util.pm in @INC (you may need to install the PMP::util module) (@INC contains: /.autofs/home/antoine/anaconda2/envs/testCustomChannel/lib/site_perl/5.26.2/x86_64-linux-thread-multi /.autofs/home/antoine/anaconda2/envs/testCustomChannel/lib/site_perl/5.26.2 /.autofs/home/antoine/anaconda2/envs/testCustomChannel/lib/5.26.2/x86_64-linux-thread-multi /.autofs/home/antoine/anaconda2/envs/testCustomChannel/lib/5.26.2 .)
As you can see it seems that some module of my Perl module are not recognized when I execute Perl. I know that to fix this issue I can modify the @INC variable and add the bin/ to the PATH and the lib/ to the PERL5LIB but I need to automatize this process during the installation of the module.
I don't really know where I should modify the environment variable. During the creation of the package by adding something in the build.sh for example ? Or should I manage that during the installation and if so, how could I do that ?
Any suggestions would be greatly appreciated,
Thanks
Edit :
meta.yaml =>
{% set name = "module_name" %}
{% set version = "0.8.3" %}
package:
name: "{{ name }}"
version: "{{ version }}"
source:
git_url: ssh://git@adress/bspcore/perl_module.git
build:
number: 0
requirements:
host:
- perl
- perl-extutils-makemaker
run:
- perl
about:
home: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
license: xxx
license_family: xxx
summary: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Build.sh =>
#!/bin/bash
if [ -f Build.PL ]; then
perl Build.PL
perl ./Build
# Make sure this goes in site
perl ./Build install --installdirs site
elif [ -f Makefile.PL ]; then
# Make sure this goes in site
perl Makefile.PL INSTALLDIRS=site
make
make install
else
echo 'Unable to find Build.PL or Makefile.PL. You need to modify build.sh.'
exit 1
fi
chmod u+rwx $PREFIX/bin/*
echo "${PREFIX}"
Edit 2 :
Another edit that could help you guys understand better my situation. I just realized that when I build the package the lib folder of my perl module in which I have PMP::util lives under lib/site_perl/5.26.0/Perl_Module. I'm pretty sure that if I'm able to install it directly under the lib/ folder it will resolve this issue. However I'm not sure how to modify the build.sh file to modify the place where we build the perl module.
回答1:
Here is a simple example of how to create a conda package that installs a Perl module (which depends on a CPAN module) that might help you solve your issue:
Installing miniconda on Linux
$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
$ bash Miniconda3-latest-Linux-x86_64.sh
# NOTE: I answered "yes" on the question:
# "Do you wish the installer to initialize Miniconda3 ?" in the
# previous command. This will modify ~/.bashrc
$ source ~/.bashrc # activates base environment
# Next command: Do not automatically activate conda for every terminal window,
# instead run "conda activate" from a given terminal window to
# activate locally. The following command also creates ~/.condarc
$ conda config --set auto_activate_base False
Creating the package:
perl-hello/meta.yaml:
package:
name: perl-hello3
version: "1.0"
source:
path: ../src # NOTE: if you had put "src" in same folder as "meta.yaml",
# conda-build would have include the src folder in info/recipe in
# the generated package. It is not necessary to include the
# source code in the generated package.
requirements:
build:
- perl >= 5.22
- make
run:
- perl >= 5.22
about:
license: Artistic
summary: Simple perl function
../src/:
$ tree ../src
../src
├── lib
│ └── My
│ └── Module.pm
└── Makefile.PL
../src/Makefile.PL:
use utf8;
use ExtUtils::MakeMaker;
WriteMakefile(
MIN_PERL_VERSION => 5.022000,
NAME => 'My::Module',
VERSION_FROM => 'lib/My/Module.pm',
PREREQ_PM =>
{
'ExtUtils::MakeMaker' => '7.12',
'Data::Dump' => 0,
},
ABSTRACT_FROM => 'lib/My/Module.pm',
AUTHOR => 'Håkon Hægland <hakon.hagland@gmail.com>',
LICENSE => 'perl',
);
../src/lib/My/Module.pm:
package My::Module;
our $VERSION = 0.01;
use strict;
use warnings;
use Exporter qw(import);
our @EXPORT = qw(hello);
our @EXPORT_OK = @EXPORT;
use Data::Dump;
sub hello {
print "Hello world!\n";
my $str = "Testing Perl module Data::Dump";
dd $str;
}
1;
build.sh:
# PERL_MM_USE_DEFAULT=1 -> automatically answer "yes" on config questions
PERL_MM_USE_DEFAULT=1 cpan App::cpanminus
perl ${PREFIX}/bin/cpanm Data::Dump
perl Makefile.PL INSTALLDIRS=site
make
make install
Note that I ran cpanm using perl ${PREFIX}/bin/cpanm. I was not able to simply run it as cpanm, see Can you rely on the shebang of an installed command during build? for more information.
Build the package
$ conda-build .
(Take a note of the generated output, and determine the path of the generated package. In my case the path name was:
/home/hakon/miniconda3/conda-bld/linux-64/perl-hello3-1.0-pl526_0.tar.bz2
Uploading the package to the anaconda server
Register a new user at Anaconda Cloud
Install the client
$ conda install anaconda-clientLogin to your account:
$ anaconda loginUpload the generated package:
$ anaconda upload /home/hakon/miniconda3/conda-bld/linux-64/perl-hello3-1.0-pl526_0.tar.bz2
Testing the package (can be done from any linux machine):
Create a new enviroment:
$ conda create --name perltest $ conda activate perltestInstall the package in the new environment:
$ conda install -c hakonhagland perl-hello3 # Alternatively: You can test the package locally before uploading with # "conda install --use-local perl-hello3"Test the package:
$ perl -E 'use My::Module; hello' Hello world! "Testing Perl module Data::Dump"
来源：https://stackoverflow.com/questions/56875121/updating-the-inc-variable-during-installation-of-a-conda-package