New submission from Jason R. Coombs:
On Python 2.7 and 3.3, if the package_data glob happens to match a directory,
it will trigger this error during build:
error: can't copy '<package>/<dirname>': doesn't exist or not a regular file
It seems that package_data is not very smart about filtering out directories,
and assumes every name matched in the glob is a file. This is particularly
inconvenient when one has a directory structure of package data. Consider:
package_data={
'bug_pkg': (
[
'html/*.*',
'html/something-1.0/*.*',
]
),
},
with a directory structure of:
.
│ setup.py
│
└───bug_pkg
│ __init__.py
│
└───html
│ index.html
│
└───something-1.0
index.dat
Since 'html/*.*' matches 'something-1.0', distutils assumes something-1.0 is a
file and tries to copy it and fails with:
error: can't copy 'bug_pkg/html/something-1.0': doesn't exist or not a regular
file
I believe distutils should be filtering out folders from package_data. In the
past, users (including myself) have worked around the issue by using '*.*' to
match only files, but that's a poor heuristic is the above example demonstrates.
This issue was encountered when using sphinx-bootstrap-theme, which adds
directories to sphinx HTML docs with directories like 'bootstrap-3.0.0', which
are difficult to not match in a glob.
Is there any reason why globs specified in package_data should not exclude all
directories?
----------
assignee: eric.araujo
components: Distutils
messages: 200298
nosy: eric.araujo, jason.coombs, tarek
priority: normal
severity: normal
status: open
title: error: can't copy '<dirname>': doesn't exist or not a regular file
versions: Python 2.7, Python 3.3, Python 3.4
_______________________________________
Python tracker <[email protected]>
<http://bugs.python.org/issue19286>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
