unstructured module (dev) doesn't work on arch
liquidsec opened this issue · comments
The unstructured module tries to install several OS dependencies that do not work on arch linux.
In several other cases, we've needed conditional logic to support arch using different package names, installing from source, etc.
the packages in question:
["libmagic-dev", "poppler-utils", "tesseract-ocr", "libreoffice", "pandoc"]
Of these, at least libmagic-dev and poppler-utils are definitely not supported.
Relevant: #1467 (comment)
Hey,
It looks like the equivalent package on arch for libmagic-dev is file and the one for poppler-utils is poppler. Arch has its own equivalents aswell. We could do something like this:
deps_ansible = [
{
"name": "Install Deps (Debian/Ubuntu)",
"package": {"name": ["libmagic-dev", "poppler-utils", "tesseract-ocr", "libreoffice", "pandoc"], "state": "present"},
"become": True,
"when": "ansible_facts['os_family'] == 'Debian'",
},
{
"name": "Install Deps (Arch)",
"package": {"name": ["file", "poppler", "tesseract", "libreoffice", "pandoc"], "state": "present"},
"become": True,
"when": "ansible_facts['os_family'] == 'Archlinux'",
},
{
"name": "Install Deps (Fedora)",
"package": {"name": ["file-devel", "poppler-utils", "tesseract", "libreoffice", "pandoc"], "state": "present"},
"become": True,
"when": "ansible_facts['os_family'] == 'Fedora'",
},
]Although Im not sure how confident I am in this solution as the developer does not state how to install unstructured dependencies on archlinux / fedora-latest / gentoo or alpine...
Yes, that looks good. Up to this point, any time I've needed to test something on a specific distro, I've used docker:
docker run --rm -it archlinuxVery soon I want to set up tests for each of the main distro families, so we won't have to worry about testing this kind of thing manually.
@domwhewell-sage if you end up making a PR, can you fork from this branch? That will hopefully show us which distros are passing.