Bio::Graphics installation steps for Rocky Linux 9 or other EL-like distros

[ernstki]

I'll be closing this issue as soon as opening it, because it's not an issue, per se. It's just to give others something to search for in the issue tracker, and so I have something to link to from our internal documentation.

I've been using Perl for quite a while, but when I cpanm Something::Something and it fails, I'm usually as mystified as anyone. So I'm also hoping this can help someone else who's in the same situation.

Steps required to get Bio::Graphics installed on Rocky Linux 9

These steps will probably work with other RHEL 9-like distros, and other related Bio::Whatever packages, too.

Note

Debian and Ubuntu already include bioperl in their default repositories, so if you're using one of those, or a derivative, installing that package from the distro repositories will save a lot of time.

Install Perl and other required development packages, to preclude build failures in the cpanm command below:

yum update
yum install perl gd-devel libxml2-devel expat-devel perl-App-cpanminus

Note

The installation of cpanminus (cpanm) from the distro repositories is actually not required, but it's helpful because it isn't obvious from cpan's output that some modules failed to build.

It's hard to even tell that cpan failed, except if you scroll up, or inspect the return code.

Next, install Bio::Graphics and look for any failed builds:

cpanm Bio::Graphics

Note

If you run cpanm as root (e.g., in a Docker container), the above actually breaks cpanm! This is why you don't normally do this stuff as root.

Since I'm installing this in a Docker container for testing, I'll just power through it by installing a necessary package from the distro repositories, so we're able to proceed:

yum install -y perl-LWP-Protocol-https

Sorting through the build logs for cpanm Bio::Graphics from above, you'll find mention that Bio::DB::SeqFeature::Store wasn't installed¹, so install that manually:

cpanm Bio::DB::SeqFeature::Store

# finally, try again
cpanm Bio::Graphics

Test to make sure the desired package actually got installed:

# perl -MBio::Graphics -e'print $INC{"Bio/Graphics.pm"} . "\n"'
/usr/local/share/perl5/5.32/Bio/Graphics.pm

Notes

• There are probably at least a handful more distro packages you could install to speed up the process relative to getting them from CPAN and running all the tests (DBI, Test::*, Capture::Tiny, XML::LibXML). The ones I chose from the distro repositories had extra dev dependencies which I wanted to avoid installing.

References

• Fix test suite with libxml2 2.13.0 shlomif/perl-XML-LibXML#87 - the patch here can be used to patch XML::LibXML to pass all its tests when using the distro-supplied libxml or libxml2, if not using a distro-supplied package and not skipping tests
• https://stackoverflow.com/a/78965715 - included the helpful suggestion of installing the distro's libxml library plus development headers preemptively

Footnotes

1. …even though it seems like cpanm should've been able to figure this out that this was a dependency on its own ¯\_(ツ)_/¯ ↩

[carandraug]

Please do not install this way. Instead, install with cpanm the modules you do use and let it figure out which module distributions to install.

Bio::Graphics is a Perl module. Bio-Graphics is a perl module distribution. When you do cpanm Bio::Graphics you are telling cpanm to install the distribution that has the Bio::Graphics module (which is `Bio-Graphics). This detail is an important distinction. When a module moves between distributions, this sort of install instructions beak.

The Bio::Graphics module provides documentation only there is no code on it. You probably do not have code that use Bio::Graphics (if you, that's an unnecessary use). Instead, you probably have code that makes use of modules that are part of the same distribution as the Bio::Graphics module.

Please, list the install you actually use with cpanm and let cpanm figure out which distributions to install. For example, run cpanm Bio::Graphics::Glyph::whiskerplot if that is what you use and it will install the Bio-Graphics distribution (or at least it will while is the correct thing to do but it will not break if that module moves into a new Bio-Graphics-Glyph distribution0.

I don't quite understand why this is causing so many errors. If a code does use Bio::Graphics::Glyph::whiskerplot, installing with cpanm Bio::Graphics requires knowing that Bio::Graphics is in the same distribution as Bio::Graphics::Glyph::whiskerplot. Simply doing cpanm Bio::Graphics::Glyph::whiskerplot requires no extra knowledge of where modules are.

[ernstki]

@carandraug I'm glad you chimed in. Agreed, in general use whatever installation method the project's maintainers wish you to use and will support, and it sounds like you're in favor of cpanm all the way.

I tried on three different current, supported Linux distros using the distro-packaged perls just to convince myself that the troubles I had weren't distro-specific. I'm not bummed about it; modern software is complicated, with long chains of transitive dependencies that can also suffer their own breakages from time to time. It's amazing it all works as well as it does!

[ernstki]

You probably do not have code that use Bio::Graphics (if you, that's an unnecessary use). Instead, you probably have code that makes use of modules that are part of the same distribution as the Bio::Graphics module.

Example #3 from the Bio::Graphics HOWTO uses use Bio::Graphics, and our script appears to be a lightly-modified version of that. I'm not the original author, just trying to make it work in a new environment.

Whether I do use Bio::Graphics or Bio::Graphics::Panel (the actual module needed) doesn't affect script operation, obviously. But the outcome of cpanm Bio::Graphics versus cpanm Bio::Graphics::Panel is basically the same; either one requires the step above to perform a successful install of the module(s) needed on a RHEL-like distro.