scripts/version.py: Use regex grouping to extract the version

The `lstrip` and `rstrip` functions take a set of characters to remove, not a
prefix/suffix. Thus `rstrip('-x86_64')` will remove any trailing characters in
the string `'-x86_64'` in any order (in effect it strips the suffix matching
the regex `[-_x468]*`). So with `18.09.4` it removes the `4` suffix resulting
in trying to `int('')` later on:

    Traceback (most recent call last):
      File "/src/scripts/versions.py", line 80, in <module>
        main()
      File "/src/scripts/versions.py", line 73, in main
        versions, reverse=True, key=operator.attrgetter('order')
      File "/src/scripts/versions.py", line 52, in order
        return (int(self.major), int(self.minor), int(self.patch)) + stage
    ValueError: invalid literal for int() with base 10: ''

Since we no longer need to check for the arch suffix (since it no longer
appears in the URLs we are traversing) we could just drop the `rstrip` and
invent a local prefix stripping helper to replace `lstrip('docker-')`. Instead
lets take advantage of the behaviour of `re.findall` which is that if the regex
contains a single `()` match that will be returned. This lets us match exactly
the sub-section of the regex we require.

While editing the regex, also ensure that the suffix is precisely `.tgz` and
not merely `tgz` by adding an explicit `\.`, previously the literal `.` would
be swallowed by the `.*` instead.

Signed-off-by: Ian Campbell <ijc@docker.com>

1 files changed, 3 insertions, 7 deletions

diff --git a/scripts/versions.py b/scripts/versions.py
index 7ad1d56..93fe0d7 100644
--- a/scripts/versions.py
+++ b/scripts/versions.py
@@ -62,13 +62,9 @@ def main():
     for url in [base_url.format(cat) for cat in categories]:
         res = requests.get(url)
         content = res.text
-        versions = [
-            Version.parse(
-                v.strip('"').lstrip('docker-').rstrip('.tgz').rstrip('-x86_64')
-            ) for v in re.findall(
-                r'"docker-[0-9]+\.[0-9]+\.[0-9]+-?.*tgz"', content
-            )
-        ]
+        versions = [Version.parse(v) for v in re.findall(
+                r'"docker-([0-9]+\.[0-9]+\.[0-9]+)-?.*tgz"', content
+        )]
         sorted_versions = sorted(
             versions, reverse=True, key=operator.attrgetter('order')
         )